Corrosion by Domestic Waters - Illinois State Water Survey · 2005-07-25 · BULLETIN 59 Corrosion by Domestic Waters by T. E. LARSON Reference: Larson, T. E. Corrosion by Domestic

ISWS-75-BUL59

BULLETIN 59

STATE OF ILLINOIS

DEPARTMENT OF REGISTRATION AND EDUCATION

Corrosion by Domestic Watersby T. E. LARSON

ILLINOIS STATE WATER SURVEY

URBANA1975

BULLETIN 59

Corrosion by Domestic Watersby T. E. LARSON

Reference: Larson, T. E. Corrosion by Domestic Waters. Illinois State Water Survey,Urbana, Bulletin 59, 1975.

Title: Corrosion by Domestic Waters.

Abstract: Essential data on corrosion gathered by the Illinois State Water Survey inisolated or programmed studies, and from experience at state institutions since 1950,are summarized. A brief review of basic fundamentals of corrosion is presented asbackground for the summaries. Also included are some of the general and specificrecommendations concerning inhibitors and construction materials that were developedthrough laboratory and field evaluations for use by architects, engineers, and institu-tional maintenance personnel. Appendixes contain a discussion of corrosion in waterwells and pumps and two ancillary papers for orientation and recognition of other fac-tors related to distribution system water quality.

Indexing Terms: Construction materials, corrosion, corrosion control, distributionsystems, laboratory tests, metals, water chemistry, water quality.

STATE OF ILLINOISHON. DANIEL WALKER, Governor

DEPARTMENT OF REGISTRATION AND EDUCATIONRONALD E. STACKLER, J.D., Director

BOARD OF NATURAL RESOURCES AND CONSERVATION

Ronald E. Stackler, J.D., Chairman

Robert H. Anderson, B.S., Engineering

Thomas Park, Ph.D., Biology

Charles E. Olmsted, Ph.D., Botany

Laurence L. Sloss, Ph.D., Geology

H. S. Gutowsky, Ph.D., Chemistry

William L. Everitt, E.E., Ph.D.,University of Illinois

John C. Guyon, Ph.D.,Southern Illinois University

STATE WATER SURVEY DIVISIONWILLIAM C. ACKERMANN, D.Sc., Chief

URBANA1975

Printed by authority of the State of Illinois–Ch. 127, IRS, Par. 58.29(6-75-1500)

CONTENTS

PAGE

Introduction . . . . . . . . . . . . . . . . . . . . . . . 1

Acknowledgments . . . . . . . . . . . . . . . . . . . . 1

Fundamentals of corrosion . . . . . . . . . . . . . . . . . . . 2

Types of corrosion . . . . . . . . . . . . . . . . . . . 2

Principles . . . . . . . . . . . . . . . . . . . . . . . 4

Effect of dissolved oxygen . . . . . . . . . . . . . . . . 5

Minerals and the electrolytic cell . . . . . . . . . . . . . . 5

Effect of pH . . . . . . . . . . . . . . . . . . . . . . 7

Effect of flow rate . . . . . . . . . . . . . . . . . . . . 7

CaCO saturation index . . . . . . . . . . . . . . . . . 93

Effects of corrosion . . . . . . . . . . . . . . . . . . . . . 10

Rusty water . . . . . . . . . . . . . . . . . . . . . . 11

Tuberculation and loss in carrying capacity . . . . . . . . . . . 12

Chemical treatment . . . . . . . . . . . . . . . . . . . . 12

Water Survey laboratory studies on corrosion . . . . . . . . . . . 13

Qualitative studies on ion migration . . . . . . . . . . . . . 13

Immersion tests with steel specimens . . . . . . . . . . . . . 14

Significance of corrosion rate measurement . . . . . . . . . . .

Mild steel . . . . . . . . . . . . . . . . . . . . . . . 20

Cast iron . . . . . . . . . . . . . . . . . . . . . . . 22

Summary . . . . . . . . . . . . . . . . . . . . . . . 28

Corrosion problems in household and institutional facilities . . . . . . . 28

Materials of construction . . . . . . . . . . . . . . . . . . 28

Water quality considerations . . . . . . . . . . . . . . . . 30

Cold water corrosion and treatment . . . . . . . . . . . .

Hot water corrosion and treatment . . . . . . . . . . . . .

Validation of performance . . . . . . . . . . . . . . . . 32

Research needs . . . . . . . . . . . . . . . . . . . . . 32

References . . . . . . . . . . . . . . . . . . . . . . . 34

Appendix A. Corrosion in water wells and pumps . . . . . . . . . . 36

Appendix B. Deterioration of water quality in distribution systems . . . . 38

Appendix C. Bacteria, corrosion, and red water . . . . . . . . . . 43

30

30

20

Corrosion by Domestic Waters

by T. E. Larson

INTRODUCTION

It is the purpose of this publication to summarize theessential data on corrosion by domestic waters gathered bythe Illinois State Water Survey in isolated studies and pro-grammed research, as well as knowledge gained from ex-perience at state institutions since 1950. Domestic water ishere defined as any water provided for human consumptionand allied uses, in contrast to sea water, brines, or conven-tional sewage effluents.

In 1939, a questionnaire was mailed to 470 Illinois wa-ter supply plants that had been in existence for over 5 yearsto assess the magnitude of corrosion problems. Repliesfrom 191 superintendents showed that 41 percent of thesewater supply systems had corrosion problems involving wa-ter mains, service lines, elevated tanks, and well pumps.

Beginning in 1943, field studies on deep well turbinepumps that had been removed for repair were made over aperiod of 6 years and resulted in 6 publications. A sum-mary of the fundamental findings is included as appendixA. With the subsequent development of reliable submers-ible pumps, those problems are now of minor importance.

In 1948, a study of corrosion of steel at low flow veloc-ities laid the groundwork for subsequent studies.

In 1949, the Water Survey was requested to preparespecifications for chemicals used in some of the state in-stitutions and to supervise the application of these chem-icals. The purpose for the specifications was to permitrealistic competitive bidding, which was formerly impos-sible with proprietary formulations. This led to the un-covering of costly problems of corrosion and maintenanceat the institutions, i.e., scale and corrosion in both low andhigh pressure boilers, condensate lines, cooling towers, dis-tribution systems, hot water systems, and a host of ancillaryproblems with swimming pools, faucets, valves, and laundryfacilities.

Since that time, corrective measures have been taken andmeans for assessment of their effectiveness have been devel-oped. Laboratory and field evaluations of inhibitors andconstruction materials have provided a breadth of experi-ence and knowledge which has permitted the developmentof general and specific recommendations to architects,engineers, and maintenance personnel at the institutions.Where possible, those recommendations related to dis-tribution systems have been included in this text.

In 1953, the American Water Works Association re-

quested the Water Survey to conduct a study on loss incarrying capacity in distribution systems. This was sup-ported in part by grants RG 4007 and WP 132 from theNational Institutes of Health, U. S. Public Health Service,from May 1, 1954, to September 30, 1968, resulting direct-ly in seven publications and indirectly in five others. Sum-maries of the fundamental findings related to corrosion ofcast iron are included, generally in sequence with time, as apart of this publication. In all, about 30 years of corrosionstudies contribute to this summary of findings. Since 1968other studies have been made as time and financial supporthas been available.

Beyond the laboratory and field research summarized inthis report is the obvious need for summaries of basicfundamentals of corrosion that have been accumulated oversome 50 to 60 years. It is readily conceded that a greatnumber of authors have contributed to this literature. Thispublication does not pretend to include a complete reviewof the literature, and only a few key references are used inthe text on principles. These are felt to be necessary to pre-sent additional background for the summaries of the studiesand to round out the presentation.

Finally, two other ancillary papers are included as ap-pendixes for orientation and recognition of other factorsrelated to distribution system water quality. Appendix Bdescribes the dynamic nature of distribution systems andthe problems that can exist. Appendix C documents actualchanges in water quality resulting from microbiologicalgrowths to the point of removal of all oxygen, and then fur-ther changes under anaerobic conditions in a 6500-footisolated water line at the end of the system.

Acknowledgments

Grateful acknowledgment is made for the administrativesupport for these studies by Dr. A. M. Buswell and Dr.William C. Ackermann, former Chief and current Chief ofthe Water Survey, respectively. Co-participants from theWater Survey (in chronological order) who contributedmuch to the findings presented herein are: John Grench,Wesley Walters, Mary C. Brooks, Donna Winn, R. M. King,R. V. Skold, E. Savinelli, Donal E. Etter, L. M. Henley, R.W. Lane, F. W. Sollo, Jr., H. W. Humphreys, and J. C. Neill.

Cast iron specimens were provided by the U. S. Pipe and

1

Foundry Company and special apparatus was contributed Frank, A. E. Griffin, Leo Louis, W. E. MacDonald, S. T.by the American Cast Iron Pipe Company. Steel specimens Powell, M. J. Taras, and E. F. Wagner.were provided by the U. S. Steel Company. Advice and Mrs. Patricia A. Motherway edited the report, Mrs. Suzisupport was provided by the AWWA Advisory Committee, O’Connor typed the final manuscript, and John Brother,consisting of E. S. Cole, E. E. Erickson, Ed Farmer, J. A. Jr., and William Motherway, Jr., prepared the illustrations.

FUNDAMENTALS OF CORROSION

Corrosion may be defined as the destruction of a metalusually by chemical or electrochemical reaction with its en-vironment. 1 Whether it be in the atmosphere, immersed inwater, or buried in the ground, a metal may be subject tochemical interaction with environmental elements.

Most metals are derived from natural mineral ores, andwhen exposed to the elements of the environment, theyhave an inherent tendency to revert to the stable forms inwhich they were originally found in the earth. A commonexample of this tendency is the rapid formation of ironrust when an unprotected common steel surface is in con-tact with water or humid air. Rust may be in the form ofany of the native oxide forms in iron ore.

Different metals and alloys have greater or lesser tenden-cies to corrode, or to revert to their natural forms. Thistendency is expressed by the relative electrochemical poten-tial related to a hydrogen electrode or to other standardizedelectrodes in a given water environment. The potential is

Table 1. Galvanic Series of Metals and Alloys

Corroded end (anodic, orleast noble)

MagnesiumMagnesium alloys

Zinc

Aluminum 2S

Cadmium

Aluminum 17ST

Inconel (active)

Hastelloy AHastelloy B

BrassesCopperBronzesCopper-nickel alloysTitaniumMonel

Steel or ironCast iron

Silver solder

Chromium-iron (active)Nickel (passive)Inconel (passive)

Ni-Resist* Chromium-iron (passive)18-8 Cr-Ni-Fe (active) 18-8 Cr-Ni-Fe (passive)18-8-3 Cr-Ni-Mo-Fe (active) 18-8-3 Cr-Ni-Mo-Fe (passive)

Hastelloy C Silver

Lead-tin solders GraphiteLeadTin

GoldProtected end (cathodic, or

Nickel (active) most noble)

* Registered U. S. Patent Office

a measure of the oxidation and reducing characteristics ofthe interface between the metal and the water, and the rela-tive electrochemical potentials of two different metals deter-mine which metal has the greater tendency to corrode.

However, the rate of corrosion is dependent largely onthe inhibition or resistance to continuing progress of thereaction by the products of corrosion in the particular wa-ter environment. This inhibition, which controls the rate,is dependent on the metal itself and a variety of conditionsinvolving the ionic components in the water as well as dis-solved oxygen, carbon dioxide, temperature, and configura-tion of the structure and movement of the water.

A generalized galvanic series of metals and alloys1 hasbeen developed and is shown in table 1. It has not beenpractical to tabulate voltage values, because these will varywith the water environment, but in general the groupedmetals have similar potentials. This galvanic series indicatesthat the least noble metals have the greatest tendency tocorrode, and the most noble metals have the least tendencyto corrode. When metals from different groups are coupledand exposed to the water environment, the more noblemetal is cathodic to the less noble metal which then cor-rodes and protects the more noble metal. This is called gal-vanic action. The farther apart the metals stand in theseries the greater the galvanic tendency. The rate of corro-sion increases with an increasing relative surface area of thecathodic metal to the anodic area.

Types of Corrosion

Each type of corrosion has its own characteristics andoffers clues to the causes and sometimes to the characteris-tics of the environment. Possible types of control may alsobe deduced from the clues.

Uniform corrosion is recognized as taking place at agenerally equal rate over the surface. The loss in weight isdirectly proportional to the time of exposure and the rateof corrosion is constant. This type of corrosion is usuallyassociated with acids or with water having a very low pHand very few protective properties. Mild steel in neutral,low-calcium, and low-alkalinity salt water corrodes at arather uniform rate.

When the corrosion rate is uniform it is possible to deter-

2

mine the weight loss over a time interval and relate this tothe electrochemical equivalent as follows:

Weight loss/time = k i

where k is the electrochemical equivalent and i is the totalcorrosion current. Although the corrosion current cannotbe measured, it may be estimated from the equation whenthe weight loss per time is known.

Pitting corrosion is nonuniform and more generally ob-served than uniform corrosion. It occurs in an environmentwhich offers some but not complete protection. The pitdevelops at a localized anodic point on the surface and con-tinues by virtue of a large cathodic area surrounding theanode. Chloride ions are particularly notorious for theirassociation with this type of corrosion of steel. Even stain-less steel is subject to pitting corrosion with relatively highchloride solutions. Pits may be sharp and deep or shallowand broad, and can occur without chlorides. In water con-taining dissolved oxygen, the oxide corrosion products de-posit over the site of the pitting action and form tubercules.

Pitting corrosion may also be associated with galvaniccorrosion, concentration-cell corrosion, and crevice corro-sion, particularly during low flow or stagnant conditions.

Galvanic corrosion is associated with the contact of twodifferent metals or alloys in the same environment. Almostall metals and substances have different solution potentials,whether in the same or in different environments. Whentwo metals come together, the difference in potential re-sults in current flow, and one of the metals becomes anodicto the other, which serves as the cathode (see table 1). Theanodic metal corrodes and the cathodic metal does not (orif so at a relatively low rate). The cathodic metal is said tobe protected at the expense of the anodic metal. To beexact, all corrosion is galvanic in the sense that an electro-chemical cell is the source of the corrosion current.

The rate of galvanic corrosion is increased by greater dif-ferences in potential between the two metals. It is in-creased by large areas of cathode relative to the area of theanode. It is generally increased by closeness of the twometals and also by increased mineralization or conductivityof the water.

Galvanic corrosion is often a great source of difficultywhere brass, bronze, or copper is in direct contact withaluminum, galvanized iron, or iron. Copper-bearing metalsare cathodic to aluminum, zinc, and iron, and their under-water contact very often results in corrosion of the lattermetals. Similarly, mill scale on steel is cathodic to thesteel; iron oxide is cathodic to iron; cement is cathodic tocopper; carbon is cathodic to iron; iron is cathodic toaluminum and to zinc.

Galvanized (zinc-coated) steel is usually more serviceablethan steel alone, because the iron exposed at joints andholidays is protected at the expense of the zinc. In general,longer life may be expected with greater thickness of the

zinc coating. Zinc in many natural waters containing alka-linity of 50 to 100 mg/l or more will form an insoluble pro-tective coating of basic zinc carbonate in the pH range of7.5 to 8.5 at room temperature or lower. In hot-watertanks, there is some evidence that zinc becomes cathodic toiron at temperatures above 140 to 160°F with certain typesof waters. Traces of copper (0.1 mg/l) in the water can‘plate out’ on zinc or iron and result in local pitting. Waterscontaining copper have a similar, more serious, effect onaluminum.

Concentration-cell corrosion is perhaps the most preva-lent type of corrosion, and because it is difficult to ascer-tain by field measurement, it is usually deduced by infer-ence. This type of corrosion occurs when there are dif-ferences in the total or the type of mineralization of theenvironment. Differences in acidity (pH), metal-ion concen-tration, anion concentration, or dissolved oxygen cause dif-ferences in the solution potential of the same metal. Dif-ferences in temperature can also induce differences in thesolution potential of the same metal.

It has been noted (page 6) that in water containing dis-solved oxygen, the corrosion products deposit at the anode,and in the secondary reaction of oxidation of ferrous ironto ferric iron and subsequent hydrolysis, hydrogen ions areformed. This greater acidity at the anode results in a hy-drogen-ion concentration cell at this point, and increasesthe rate of corrosion. In the same instance, dissolved oxy-gen cannot diffuse or penetrate to the anode surface be-cause it first reacts with the ferrous iron; thus, there is anabsence of oxygen at the anode. But oxygen can diffuse tothe cathode area and result in an oxygen concentration cellwhich also increases the rate of corrosion at the point ofabsence of oxygen. Hydroxide ions accumulate at thecathode area, resulting in drastic reduction in hydrogen-ionconcentration, which therefore enhances the concentrationcell related to the development of hydrogen ions at theanode.

It should be noted that although the dissolved oxygenusually stimulates corrosion, the loss in metal takes place atthe anode, where there is no dissolved oxygen.

Oxygen concentration cells develop at water surfaces ex-posed to air, accelerating corrosion a short distance belowthe surface. The dissolved oxygen concentration is replacedby diffusion from air and remains high at and near the sur-face, but does not replenish as rapidly at lower depths be-cause of the distance for diffusion. Therefore, the corrosiontakes place at a level slightly below the surface rather thanat the surface.

Dirt and debris or local chemical precipitates on a metalsurface hinder oxygen diffusion by covering the metal atlocal areas. Thus, corrosion takes place under the deposit.

Thus it is evident that any nonadherent deposition onmetal can start a chain of circumstances which result in anoxygen concentration cell.

3

Crevice corrosion might be classed as a form of concen-tration-cell corrosion because when oxygen is spent on cor-rosion in a crevice, it is difficult for more oxygen to reachthe metal by diffusion into the depths of the crevice. Thecrack, crevice, or uneven joint between two surfaces of thesame metal bound together face to face behaves as a pitwhere oxygen can reach the exposed surface but becomesdeficient in the crevice, thus forming an oxygen concentra-tion cell where corrosion takes place.

Dezincification is the result of removal of zinc from itsalloy with copper (brass). Copper remains at the surface ofthe brass as the zinc is dissolved. Soft unstable waters, es-pecially those with a high CO

2content, are particularly

aggressive to Muntz metal and yellow brass. Red brass andAdmiralty metal are more resistant. The occurrence ofplug-type dezincification and dezincification at threadedjoints suggests that debris and crevices may initiate oxygenconcentration cells and result in dezincification.

Graphitization is a form of corrosion of cast iron inhighly mineralized water or waters with a low pH, whichresults in the removal of the iron silicon metal alloy makingup one of the phases of the cast iron microstructure. Ablack, spongy-appearing, but hard mass of graphite remains.The graphite dispersed in the cast iron serves as the cathodefor a large number of small galvanic cells, and the iron-silicon alloy becomes the anode.

Cast-iron products derive their shape from the freezingof a liquid alloy in a mold; thus solidification takes placeunder different conditions over the cross section of anycast shape. The portion near the mold wall may have adifferent composition, freezing rate, and structure than thematerial farther from the mold wall. There is likely to beless graphite at the surface of the casting; thus we have thefamiliar observation that machining of cast iron produces alower corrosion resistance at the machined areas.

The high corrosion resistance of cast iron in soil is prob-ably due to the formation of a silicate film at the surfacewhich together with the initial corrosion products formsan effective barrier. In mine water having a low pH andhigh concentration of iron and/or copper salts, the silicatefilm may be quite soluble. The removal of the film thencauses rapid corrosion of the metallic phases of the castiron, leaving the skeleton of graphite embedded in corro-sion products.

Stress corrosion results from tensile stress, usually of ex-ternal origin, on the metal or alloy. The corrosion usually(but not always) takes place selectively at the microstruc-ture grain boundaries in the metal. Repeated rupture of aprotective film on the surface provides a continuouslyanodic region.

Corrosion fatigue resulting from alternate stress con-ditions is usually more rapid than steady-state stress corro-sion. The alternation of the stress disturbs such protective

film as may develop at the anode site and enhances the rateof corrosion.

Erosion corrosion results from the removal of the pro-tective film of corrosion products which serves as a barrierto corrosive attack of some metals. Many metals such asaluminum, austenitic stainless steel, and passive iron arecompletely protected by a film (such as aluminum oxide onaluminum in favorable environments). The erosion, general-ly at high velocity, may take place through removal of thefilm by abrasive, suspended material. Friction between twoadjoining surfaces may also permit corrosion to continue.

Impingement attack on copper pipe is also an exampleof apparent erosion corrosion, as it is particularly evident atvelocities greater than 5 to 8 feet per second (fps) and atsudden changes in direction of flow at joints and elbows.

Cavitation corrosion is usually associated with highvelocity and sudden changes in velocity direction whichcauses gas-bubble formation at low pressure points andresolution of the gases at high pressure points. The attackoccurs downstream from the direction change caused by aconstriction such as a valve or protrusion at a joint. Carbondioxide in the gas bubbles develops an acidic film at thewater surface. If oxygen is present, its concentration at thewater surface is also greater than in the body of the waterand further accelerates corrosion.

Wire drawing in faucet seats appears as grooves acrossthe face of the seat. Apparently this is a type of cavitationcorrosion or erosion corrosion in which chloramine servesas a corrosion accelerator. Monel metal seems to be theonly alloy that resists this type of attack.

Stray currents have often been blamed for corrosionoccurrences where other causes have been present. Withdirect-current electric railways, serious corrosion problemswith underground structures have properly been traced tostray currents, but almost invariably this has been externalcorrosion of piping rather than internal. All metals havegreater conductivity than the surrounding environment, andcurrent will stay with the metal until there is a discontinui-ty. An excess of electrons will leave the metal at the pointswhere the environment is more highly conductive andclosest to another conductive receptor for the current. Cor-rosion takes place at the anode, the point where the cur-rent leaves the metal to be returned to the power source.Currents seek out and travel by the path of least resistance.

The use of insulating connections or proper counter-current applications such as cathodic protection can elimi-nate this problem. However, improper use without a satis-factory survey of the problem and advice by specialists,can result in accelerated corrosion rather than prevention.

Principles

Although destruction of metals can result from directchemical reactions, it is more generally considered as an

4

electrochemical process. In the water environment, thisprocess is usually destructive to materials of construction,but it may also present conditions conducive to protectivereactions.

It is generally accepted that corrosion results from theflow of electric current between electrodes or anodic andcathodic areas on the metal surface. These areas may bemicroscopic and in very close proximity, thereby causinggenerally uniform corrosion and often ‘red’ water; or theymay be large and somewhat remote from one another andcause pitting, with or without tuberculation. Potential dif-ferences may be induced by various conditions, some by thecharacteristics of the metal and some by the character ofthe water at the boundary surface or interface. Especiallysignificant are variations in the composition of the metal orthe water from point to point on the contact surface. Im-purities in the metal, sediment accumulations, adherentbacterial slimes, accumulations of the products of corro-sion – all are related either directly or indirectly to thedevelopment of anodic and cathodic areas for corrosioncircuits.

In almost all forms of pipe corrosion, the metal goes intosolution at the anode areas. Because electrons are releasedto the surface when the metal dissolves, the metal developsan electrical potential. Electrons liberated at these areasflow through the metal to the cathode areas where theybecome involved in another chemical reaction and the metaldevelops another electrical potential. Control of corrosionby water treatment methods aims at retarding either or bothof the primary electrode reactions.

Because electric currents are carried by ions, as con-ductors in water, and because these ions affect the poten-tials at both the anode and cathode, some discussion of thesoluble components in water is warranted.

To some extent, almost all mineral salts dissolve in wa-ter, from insignificant traces to gross concentrations ex-ceeding that of salt in sea water. On solution, these saltsseparate into two types of ions, anions and cations. Theseions have opposite electric charges and are kept apart bythe water itself. These ions are responsible for the abilityof water to conduct an electric current. Pure water hasrelatively few ions. Only one one-hundred thousandth ofone percent of water separates into hydrogen cations andhydroxyl anions. Therefore, pure water, free from mineralsalts, has very little capacity for carrying electric current.Its electric conductivity therefore is extremely low.

However, when minerals are dissolved in water, theresultant ions provide the necessary conductivity to permitthe corrosion current to flow, with negatively chargedanions moving to the anode and positively charged cationsmoving to the cathode. Their accumulation at the respec-tive electrodes is limited by other reactions which take placeat these points.

The basic electrode reactions involving the transfer ofelectrons may be represented as fellows:

Anode: Fe – 2∈→ Fe++

Cathode: 2 + 2H2O → H2 + 2OH–

When dissolved oxygen is present, the cathode reaction maybe represented as: 2 + H2O + 1/2 O2 → 2OH

– . There-

fore, with or without dissolved oxygen, the same amountof hydroxide ion is formed at the cathode and an alkalinecondition prevails.

Effect of Dissolved Oxygen

The most prevalent and at the same time the most po-tent of the common corrosive agents is dissolved oxygen.Oxygen in any form of water, be it humidity, rain, drinkingwater, or sea water, has long been recognized as a destroyerof ferrous metals. When present in public water supplies itsconcentration may range from a few mg/l to as much as 14mg/l. This is usually small in comparison with other ingre-dients. It is no more than half the concentration of nitro-gen, for example.

As the water passes through a distribution system, rarelyis more than 1 mg/l of oxygen used up in corroding themetal, except in low-circulation areas or dead ends. At a5-fps velocity in a 24-inch main, it is doubtful that a changein dissolved oxygen of more than 0.1 mg/l per mile ever oc-curs. But this is a continuous supply, a continuing feed ofthis destroyer to the pipe wall over years of service.

Were water of this same oxygen concentration used in ahome hot-water heating system largely composed of steelwith no replenishment or replacement of the water, the sys-tem would last indefinitely, because the limited supply ofoxygen would be quickly removed, and little or no corro-sion would then be possible.

The discovery of a method more economical than me-chanical deaeration for removing dissolved oxygen from wa-ter could reduce the present serious corrosion by at least90 percent. Many years ago Speller2 showed that passageof water through iron filings removed dissolved oxygen, butthis technique is hampered by the problem of replenish-ment of the filings and clogging of the filters.

Minerals and the Electrolytic Cell

If sodium chloride was the only mineral present, thenegative chloride anions would be attracted to the positiveanode where Fe++ is present, and the positive sodium cati-ons would be attracted to the negative cathode (see figure1a). Thus at the cathode there would be a sodium hydrox-ide solution, and at the anode a soluble ferrous chloridesolution. But, with dissolved oxygen present, the ferrouschloride solution is oxidized to form ferric chloride, part

5

∈

∈

Figure 1. Electrolytic corrosion of iron in a) sodium chloride solution and b) calcium bicarbonate solution

of which in turn reacts with water itself to form an insol-uble ferric hydroxide and hydrochloric acid

2FeCl2

+ 5H 2O + 1/2O2

→ 2Fe(OH)3

+ 4HCl

Therefore, at the anode there is a mixture of dilutehydrochloric acid with a solution of ferrous chloride and aprecipitate of ferric hydroxide. This is then the majorcorrosion product.

The high pH at the cathode in the presence of dissolvedoxygen and low pH at the anode increases the potential dif-ference between the electrodes, and thus increases the cor-responding current flow and the corresponding rate ofsolution or corrosion of iron.

However, if the only ions present were those of a cal-cium bicarbonate solution rather than sodium chloride, thecalcium cations would be attracted to the cathode and thebicarbonate anions to the anode (see figure 1b). Thus, atthe cathode there would be a calcium hydroxide solution,and at the anode, a ferrous bicarbonate solution, part ofwhich by reaction with dissolved oxygen forms insolubleferric hydroxide and carbonic acid. Carbonic acid is a muchweaker acid than hydrochloric acid. In fact, the combina-tion of carbonic acid and bicarbonate ion serves as a ‘buffer’

to prevent an excessive reduction in the pH at this point.At the cathode of this system the calcium hydroxide re-

acts with the calcium bicarbonate of the main body of thewater. As these mix together, they form insoluble calciumcarbonate that when deposited on the cathode surface, de-velops a thin layer of limestone and serves as an insulator.One asset of this is to resist the passage of ions and current.Another asset would be a deposit of calcium carbonatewhich resists the diffusion of oxygen to the metal sur-face. In contrast, at the anode, the ferric hydroxide,usually with some underlying magnetite, offers little resis-tance to ionic movement and current flow.

In natural waters, sodium chloride and calcium bicar-bonate seldom occur alone without the other. Various pro-portions of calcium bicarbonate and sodium chloride (andother minerals) may be present, so it is not surprising thatdifferent waters vary in their tendency to corrode and intheir capacity to hinder or prevent corrosion.

It is evident that reactions at the metal interface arecritical, and the general mineral quality of the water has adirect influence upon these reactions. High ratios of bicar-bonate to chloride and of calcium to sodium are conduciveto inhibition and protection.

6

Critical, also, is the rate of diffusion of oxygen to theinterface. Dissolved oxygen has no electrical charge, andtherefore is not ionized. It is used up, however, in the oxi-dation of the ferrous ion at the anode, and in the formationof hydroxyl ions at the cathode. These uses tend to de-plete the oxygen at both electrodes, and thereby attractoxygen from the main body of the water. This diffusionto the metal surface is slow and controlled in part by theconcentration of oxygen in the main body of the water.The greater the concentration, the greater the rate of dif-fusion to the surface.

When ferrous ions are contained at the anode by de-posits of ferric hydroxide, the oxygen that diffuses to thisarea is used up in oxidizing ferrous ions to ferric hydroxidebefore it can get to the anode surface. This surface area istherefore devoid of oxygen. At the cathode, oxygen doesget to the surface, and the rate of diffusion helps to governthe rate of corrosion when there is little or no calcium. Butwith sufficient calcium present, a tightly adherent depositof calcium carbonate resists the diffusion of oxygen to themetal surface, and the corrosion rate is reduced accordingly.

Effect of pH

In 1924 Whitman, Russell, and Altieri3 concluded fromcareful experiments with Cambridge, Massachusetts, waterthat, in the pH range 4.1 to 10.0 at 22°C and 4.3 to 9.0 at40°C, hydrogen ion concentration has no effect on the rateof corrosion, and the main variable in this pH region is therate at which dissolved oxygen diffuses to the metal surface.This conclusion was not intended to apply to waters ofother mineral character, but it has been variously misquotedor enlarged upon to imply that dissolved oxygen controlsthe rate of corrosion in natural waters. The latter inter-pretation of the Cambridge results was proved incorrect byBaylis4 in 1926. He demonstrated the practical value ofcalcium carbonate protection by controlled pH and by thelow solubility of ferrous carbonate at a pH greater than 8.

Because, in reality, each addition of caustic or acid toadjust the pH of Cambridge water produced a water of dif-ferent mineral quality, the conclusions reached by Whitmanand his colleagues need not be applicable to water of thesame mineral quality as that at Cambridge or to potablewaters in general.

Long experience has taught water works personnel thevalue of pH control for corrosion protection, but it has alsoshown that factors other than pH and dissolved oxygen in-fluence corrosion rates.

Effect of Flow Rate

Another factor controlling the rate of diffusion is thevelocity of flow of the water in pipes. With slow laminar

straight-line flow, the gradient from high concentration ofoxygen in the main body to low concentration near the sur-face may be quite flat, so that the rate of diffusion is slow(figure 2a). With increasing turbulence, the laminar-flowlayer adjacent to the pipe wall becomes thinner and thegradient steeper over the shorter path (figure 2b). There-fore, in corrosive waters, turbulence at higher flow ratespermits oxygen to reach the cathode surface more rapidly,and the corrosion rates are accordingly greater.

However, with relatively noncorrosive, balanced waterswith high ratios of calcium to sodium and high alkalinity,the higher flow rates also increase the concentration gradientfor diffusion and decrease the distance that calcium andalkalinity must travel to the cathode to form and depositcalcium carbonate. This decreases the corrosion rates. Itwould appear that the higher the ratio of calcium andalkalinity to dissolved oxygen, the greater the protectivetendencies of the water.

From corrosion prevention experience, it is well knownthat, to protect metals under conditions brought about bystagnant water, highly excessive inhibitor concentrationsmust be used. Such conditions often dictate the use ofcathodic protection for corrosion prevention.

It has been an accepted fact that, where flow rates arelow, dead ends are often the most troublesome areas in dis-tribution systems, with regard to tuberculation and redwater, and often, tastes and odors. The reasons for the dif-ficulty are the long time of contact, which causes the ac-cumulation of iron resulting from corrosion, and the factthat the protective ingredients in the water are not able toreact at the point of corrosion and on the surface of themetal because of the low diffusion gradient at idle or lowflow rates. The thickness of the laminar layer limits therate of diffusion of ions to the pipe surface.

The concept of employing mechanical means to bringthe protective ingredients to the pipe surface and thus tothe reaction products of corrosion was suggested in 1959.Larson5 and Larson and Skold6 have shown that velocity(rate of flow) enhances tuberculation at 0.2 fps and calciumcarbonate protection at 2.0 fps in Champaign-Urbana tapwater (figure 3). Baylis 7 in 1953 also showed the effect ofvelocity. McCauley 8 gives strong weight to this factor.Eliassen et a1.9 in 1956 related corrosion rate to velocityfor a water with a negative saturation index. They indicatedthat the increasing velocity decreased the path of diffusionof oxygen through the laminar flow layer adjacent to thepipe wall. The oxygen in this nonprotective water therebyaccelerated corrosion.

For a water with sufficient protective characteristics, theeffect of oxygen can be nullified, or made to contribute tothe protective action, when the thickness of the laminarlayer is reduced.5 This thickness δ is related to velocityV and arbitrarily to the friction factor f by:

7

Figure 2. Comparison of oxygen diffusion under laminar and turbulent flow

and therefore, in turn, related to the Reynolds number, is determined from the same measurements of velocity andVD/v (figure 4). D is the diameter; v, the viscosity. Be- head loss, the relationship between the Reynolds numbercause the Hazen-Williams coefficient, C, is normally used by and the Hazen-Williams coefficient is introduced to indicatewater utility investigators to establish roughness of pipe and the laminar thickness at various velocities. In water supply

8

Figure 3. Effect of velocity on tuberculation0.2 fps 1.0 fps 2.0 fps

Figure 4. Relationship of laminar flow, layer thickness, friction, and Reynolds number

practice, the range of turbulent pipe flows is usually con-sidered to be between the limits for perfectly smooth pipeand limiting velocities beyond which there is completeturbulence – that is, no change in friction factor. TheHazen-Williams coefficient has been plotted in figure 4 for avelocity of 4 fps, but there is very little difference in theposition of the curves at other velocities.

It will be noted that the velocity of flow plays an im-portant role with regard to the thickness of the laminarflow layer, which is decreased by approximately half as thevelocity is doubled. It will also be noted that this thicknessis greater for smooth pipe than for rough pipe. For in-stance, with a coefficient of 145 for the example indicatedin figure 4, the thickness would be approximately 0.32 inch,which indicates the greater need for higher velocities at newinstallations. This, in part, may explain discrepancies in ob-served tuberculation at various locations in the same com-munity .

Figure 4 does not show the effect of temperature. Tem-perature affects the viscosity v , which, in turn, may increasethe thickness of the laminar layer by approximately 30 per-cent as the temperature decreases to 35°F.

An example of corrosion inhibition by the combined ap-plication of calcium carbonate stability and velocity wasdemonstrated at a community which had suffered from red-water problems for nearly 25 years.10 The system was de-signed for fire flows greater than 500 gallons per minute(gpm), but the daily demand was only approximately

12,000 gallons per day (gpd). The maximum demand veloc-ity in the residential area was about 0.04 fps in 6-inch pipe,which is a flow rate in the laminar range of the Reynoldsnumber. Induced circulation at 60 gpm helped alleviatethe distribution system problem, but data showed thatcorrosion continued until pH adjustment to a +0.3 satura-tion index eliminated the problem. When the induced ve-locity was reduced to 30 gpm (0.4 to 0.9 fps), a saturationindex of approximately +0.6 was necessary to avoid cor-rosion by this water, which was softened to 90 mg/l hard-ness by ion exchange.

Although the principle of calcium carbonate stability isthe one most widely applied and found to be most widelyeffective in treatment for the protection of distribution sys-tems, there are many notable exceptions to this principle.A review and evaluation of this principle is warranted.

CaCO3 Saturation Index

The equation for the saturation index was derived byLangelier : 11

I = pH – pHs = pH – K + log Ca + log Alk

in which I is the saturation index; pH, the actual pH; andpHs , the pH at saturation with CaCO3 . K is the log Ks /K2(K are thermodynamic constants, corrected fors and K2ionic strength and temperature).

9

Ryznar index = 2pHs – pHor = 2K + 2(log Ca) + 2(log Alk) – pHor = H +/(Hs

+ )2 = (Ca)2 (Alk) /(K´)

Ryznar stated that the scale-forming tendency increased asthe index decreased from 6, or as the ratio of H + /(H s

+ )2 de-creased from 10 6 .

McCauley 13 proposed a driving force index (DFI):

DFI = (Ca+ +)(CO3– –

)/K s × 10 10

with which he estimates the carbonate ion concentrationfrom the alkalinity, pH, and K´2 .

All of these approaches recognize a weakness in the satu-ration index as a direct criterion of the magnitude of excesscalcium carbonate. Incidentally, Langelier never intendedit to be such a criterion, other than as a relative indicatorthat was subject to interpretation. Langelier 11 emphasizedthat the index “is an indication of directional tendency and

Then the antilog of I would be:

Hs+ /H+ = (Ca)(Alk)/(Ca – x)(Alk – x)

from Which x, the initial excess, can be calculated. Thismethod of calculation indicates a temporary excess but notthe ultimate or equilibrium excess, which is always smaller.Ryznar 12 emphasized the importance of the concentra-tions of calcium and alkalinity to scale-forming properties:

The equation may also be expressed as:

antilog I = (Ca) (Alk)/K´(H+) = Hs+ /H +

in which K´ is the antilog of K. These relationships exist:

I f I = 0 ; antilog I = 1.0I = +0.4; antilog I = 2.5I = +1.0; antilog I = 10.0

Another way of using this index is to relate the actualpH to the imaginary calcium and alkalinity concentrationsthat would exist if the actual concentrations were decreasedby equal amounts (x) until equilibrium existed with theactual pH. Then:

I = pH – pHs = K – log (Ca – x) – log (Alk – x)– K + log Ca + log Alk

of driving force, but it is in no way a measure of capacity.”Caldwell and Lawrence14 established curves from which

the equilibrium excess calcium carbonate could be deter-mined for water at temperatures of 25 and 100ºC. Thesecurves are not entirely adequate for waters at lower tem-peratures, but they are helpful because interpretations forother temperatures can be derived.

The relative excesses of calcium carbonate at 25ºC(77ºF) are shown in figure 5 for equal calcium and alka-linity concentrations.5 This is perhaps the best numericalindicator of the available calcium carbonate that may beexpected from supersaturated water and is similar to themomentary index reported by Dye.15 The reaction rates,however, are still dependent on: 1) temperature, 2) the in-fluence of OH– and H + developed at cathodic and anodicareas of the corroding surface, and 3) the buffering effectof greater concentrations of both calcium and alkalinity inthe water. This buffering effect might be considered as theability to maintain a potential or latent protective concen-tration after the deposition of a unit of calcium carbonate.These data on excess calcium carbonate may be made moreuseful by the inclusion of a measure of the buffering capac-ity related to the concentrations of calcium and alkalinity.

Experience has shown that the greatest weakness in thesaturation index occurs with waters of relatively low alka-linity and calcium. Such waters have a saturation pHgreater than 8.0 to 8.3. For some waters, the saturation in-dex may have to be +1.0 or more to establish the saturationpH for calcium carbonate. At this level the buffer capac-ity is at a minimum. At higher levels, recent literature alsoindicates the formation of complexes such as CaOH+ ,MgOH+ , CaHCO3

+ , MgHCO3+ , and soluble CaCO3 and

MgCO3 . These complexes remove calcium and alkalinity(HCO3

– , CO 3– –, and OH–) from active concentrations of

their normal form. Analytical procedures fail to distin-guish between these forms; therefore, there is a need to es-tablish equilibrium constants for these complexes. Similarconstants are required for CaHSO4

+ , MgHSO4+ , and soluble

CaSO4 and MgSO4 . In each case a temperature coefficientfor the constant is required.

EFFECTS OF CORROSION

Corrosion is a problem that has plagued water utilitiesfor as many years as metallic pipes, appurtenances, andwater-using facilities have been employed. Most annoyingand frustrating to water utility personnel, and to the con-sumer, is the deterioration of a clear, palatable, and safewater produced with care at the water treatment plant thatoccurs between the outlet and the point of use. Researchon corrosion in distribution systems has been sporadic for

over 50 years, with notable successes in developing con-cepts, in the development and proper use of materials ofconstruction, and in applying better knowledge of treat-ment to produce less aggressive water quality.

The effects of corrosion in public water supply systemsare varied and numerous. Ultimate destruction is rare, butpitting to the point of penetration can occur with poorlyprotected steel pipe. The major effects are related to dis-

10

2 2

Figure 5. Excess calcium carbonate from supersaturated water

solution of iron and to tuberculation. For the past decadeor two, cement lined water mains have been installed withincreasing frequency, because of their excellent record ofperformance. However, thousands of miles of the olderpoorly coated pipe are still in service, and these are stillsubject to the effects of corrosion.

Rusty Water

One obvious result of corrosion can be the appearance of‘red water’ at the household tap. This rusty water, result-ing from corrosion, consists of hydrated iron oxide sus-pended as particles that cause the water to be turbid andunsightly. The rusty water stains household appliances andporcelain ware, and clothes laundered in such water arealso stained. (Rusty water may also be apparent at the tapwhen the water supply source contains iron in excess. Thisis limited to some well water sources, and should not bemistaken for iron resulting from corrosion.)

Because distribution systems are designed to distributewater from one or more central sources to various areasand points of demand, the pipes are large near the sourceand become progressively smaller and interconnected towardthe farthest points of demand. The larger mains usuallycarry a relatively sustained large flow of water, and thesmallest carry erratic flows depending upon the number ofpoints of demand. Minimal limits of size are built into thedesign to permit large flows for fire protection. Therefore,rusty (red) water as a result of corrosion is a lesser problemnear the source than at the farthest points of demand. Thisis, in part, because the time of contact with the pipe perunit volume of water is relatively short for the larger mains,and longer for the smaller pipes where the demand is low ornegligible for significant periods of time. Furthermore, ascorrosion takes place and rust particles develop, they settle tothe bottom of the pipe at periods of low or negligible de-mand, and are picked up and redistributed at periods ofhigh or instant demand.

11

Tuberculation and Loss in Carrying Capacity rode the pipes of galvanized iron, copper, copper alloys, a-

Another serious problem resulting from corrosion is evi- luminum, and steel (see page 14). The choice of construc-

dent as loss in carrying capacity of the water distribution tion materials for other water uses may also be critical. Of-

system. When the corrosion products form nodules or tu- ten there are alternatives. For different qualities of water

bercules on the interior surface of the mains or pipes, these and specific uses, one metal may be less subject to corrosion

protrusions offer resistance to the flow of water. This re- than others. Each metal or alloy has its limitations depend-

sults in a decrease in flow rate, because the water pressure ing on the use and the quality of the water.

at the point of demand is reduced. This reduces or stops theflow of water at the tap in the taller buildings, and limits therate of flow to the water hydrant for fire fighting. In order Chemical Treatment

to overcome this reduced pressure or head loss, higher Because water quality is implicated in the problem ofpumping pressures must be generated at the pumping sta- corrosion, attempts have been made for over 70 years totion. Uncontrolled tuberculation can double the cost of alter the quality by chemical treatment. However, chemicalpumping. control usually must be considered as a supplement to

Cast iron pipe rarely fails structurally as the result of proper choice of materials and protective coatings, not as acorrosion, but iron can be leached out from the iron-graph- substitute. Chemical control cannot be expected to over-ite matrix, leaving the graphite in a dense porous form. come improper flow conditions, poorly designed distribu-However, the rate of leaching by corrosion decreases as the tion systems, defective materials, grossly faulty coatings,penetration proceeds. Such corrosion is called graphitiza- and under-design of copper pipe.tion, and becomes serious only with highly mineralized wa- Chemical control involves expense and constant surveil-ter which is rarely used for public water supplies. lance. The chemicals are costly, require proper feeders as

Because loss in carrying capacity is associated with the well as storage, and must be carefully selected. They can be‘roughness’ of the interior of the pipe, it is important to hazardous to personnel if handled carelessly. Analyticaldifferentiate between the different types of roughness that tests must be made at regular intervals to ensure proper ap-may develop : plication and to evaluate performance. None of these

1) The slime deposit or growth of bacteria, whether or problems are insurmountable, however, and as more isnot manganese or iron is present, can occur over the learned about chemical control, its role should become in-entire surface of the main. creasingly important.

2) Deposits of silt from untreated water is not uncom- On the whole, chemical control has been more success-mon in old mains or those operating under conditions ful with water for industry than with water for drinking.of low flow. Such old depositions are known to be In many industrial processes, chronic or acute toxicity ispresent even where water clarification has been prac- not generally a significant factor, and therefore there is noticed for over 20 years. In addition to sand, samples serious restriction in kind and quantity of chemical control.have been found to contain alumino-silicate clays and Because the water is not used for drinking or other house-microorganisms. The effect of these deposits on hold or institutional purposes, high concentrations of in-carrying capacity is probably only slightly greater than hibitors, such as chromates, nitrites, phosphates, organicsthat which calculations would indicate from the re- such as tannin, and synthetics such as mercaptans can beduction of the pipe cross section. tolerable and, when quantity requirements are low, eco-

3) Incrustation, or the formation of a crust over the nomical.metal surface, may be of various kinds, such as a) tu- For example, chromates, phosphates, and organics haveberculation in the form of nodules or spicules, re- been used in relatively large quantities with exceptional suc-sulting from localized corrosion; or b) after-precipi- cess in treatment of cooling waters cycled through heat ex-tation or uniform deposition (not always a hard changers and cooling towers. Preheaters with oxygencrust) of insoluble water constituents on the pipe scavenger chemical treatment and pH adjustment are alsowall. effective and are economic assets to boiler feedwater con-

Corrosion products also retard heat transfer for heating trol. With water for drinking, however, there are seriousor cooling water, and therefore an excessive amount of wa- limitations on kind of chemicals used, because of effectster is required to provide the heating or cooling that is on taste, odor, color, turbidity, and toxicity, and almostdesired. equally serious limitations on quantity, because a great

In addition to water distribution systems, the water used volume of water must be supplied at low cost and the wa-for domestic, industrial, and institutional purposes can cor- ter once used is not recycled.

12

WATER SURVEY LABORATORY STUDIES ON CORROSION

The literature abounds with unsubstantiated observationson corrosion without data. How often are conclusions pub-lished without data on time of exposure, stress, stray cur-rents, composition of the metal or alloy, the quality of thewater environment (pH, temperature, dissolved oxygen,mineral analysis), concentration of inhibitor, rate of flow,and a host of other variables such as trace elements, po-tential for presence of copper, galvanic action, or turbu-lence?

It is one thing to study the literature on corrosion test-ing and another to try to extrapolate the findings to applyto specific problems. Therefore, it is necessary to conductlaboratory studies under controlled conditions to evaluatethe relative effects of primary variables as a basis for ex-trapolation for specific applications. The following studieshave been directed largely to evaluate the mineral qualityvariables on corrosion, and to determine the effects anddemonstrate means for reducing corrosion.

Needless to say, research on corrosion involves manymeasurements for controlled conditions over a significantperiod of time. In one 10-month period, over 500 speci-mens were used for approximately 45 series of tests, duringwhich about 4000 operations such as cleaning and weighingspecimens and quality control tests were made. Over16,000 current and voltage readings accompanied thesetests.

Simulating the normal (real life) corrosion rate of metalsin the laboratory is important. By design, metals in nor-mal conditions are exposed for many years. The time ele-ment in laboratory studies is much shorter. It is wellknown that at first exposure, metals in laboratory experi-ments corrode at a very high rate. Depending upon variablescontributing to acceleration and inhibition, the first ex-posure rate decreases to a constant or equilibrium ratewhich may require 50 to 100 days to achieve. This equi-librium rate is the one closest to normal exposure.

At first a major problem for all corrosion studies was thenecessity for destroying the specimen to determine thecorrosion rate. This could involve many specimens to deter-mine the equilibrium rate. The problem was overcome bythe observation of Ronald Skold 18 that during polarizationtests at low applications of current to the specimens, thechange in the potential of the specimens was high for highcorrosion rates, and low for low corrosion rates. Becausethe required current density was in the microampere rangefor only a few minutes, the specimen was not destroyed,and it was possible to develop an empirical relationshipbetween the instantaneous corrosion rate and the changein potential per unit of applied current density. Within afew years after the publication of this discovery 17,18 t hetheory and improved application of this technique weredeveloped by many investigators in the field, to the point

This section on laboratory studies begins with some pre-liminary studies conducted by the Illinois State Water Sur-vey on corrosion of steel, followed by our studies with castiron, and with water quality parameters such as chloride,alkalinity, calcium, and pH.

that the application of this finding to corrosion research ex-panded at an exponential rate.

Qualitative Studies on Ion Migration

In 1948, a preliminary investigation was begun to ob-tain qualitative data on the water composition at the cath-ode and the anode, as affected by waters of differentmineral compositions.19 The investigation was designed tostudy the general water quality conditions that develop be-tween two iron electrodes under the influence of an artifi-cially impressed voltage to produce a current density of alimited magnitude. The 5×6-inch electrodes were locatedat opposite ends of an electrolysis cell divided into nine270-ml compartments by vertical, parallel, porous alundumplates (figure 6).

University of Illinois tap water was permitted to flowthrough the center compartment at a rate of 300 ml perminute, while in the remaining compartments, the water wasin a quasi-stagnant condition. No attempt was made toaerate the water or to exclude dissolved oxygen. The com-position of the water is indicated in table 2.

Progressive quality changes occurred in each compart-ment with increasing milliampere-hours of current con-

Figure 6. Experimental corrosion cell

13

Table 2. Tap Water Composition

me/l

Iron (Fe)Manganese (Mn)Calcium (Ca)Magnesium (Mg)Sodium (Na)Silica (SiO 2 )Fluoride (F)Chloride (Cl)Nitrate (NO 3)Sulfate (SO 4 )Alkalinity (as CaCO3)Hardness (as CaCO 3 )Dissolved oxygenTemperature (degrees

Fahrenheit)pH

mg/l

tracetrace60.024.046.019.0

0.36.00.29.6

330.0250.0

6.0

557.4

3.02.0

0.20

0.206.605.00

sumption. The changes were particularly significant in theend compartments at the electrodes.

The increase in pH at the cathode is shown in figure 7at three current densities (cd) 0.9, 1.7, and 3.5 milliam-peres per square foot (ma/sq ft), and the accompanyingreduction in calcium and magnesium (me/l) is shown infigures 8 and 9. At low current density, the pH was notaffected so greatly for equivalent milliampere-hour values,and, accordingly, magnesium precipitation was minimal.

The general distribution of calcium, magnesium, andalkalinity concentrations (me/l) in the various compartmentsis shown in figure 10. The considerable loss of alkalinitytoward the anode compartment was noteworthy.

In several tests, measurements were made on polyphos-phate and silica concentrations in each of the compartments.When polyphosphate or silica was present originally in allthe compartments, these tests showed a progressive decreasein the concentration of these ingredients in the compart-ments near the anode and the cathode. When all but thecenter compartments were free from polyphosphate, how-ever, the migration of polyphosphate was definitely towardthe cathode at low current density (0.4 ma/sq ft). At highcurrent density (4 ma/sq ft), polyphosphate was found tomigrate toward anode and cathode at equal rates.

Figure 7. pH at cathode Figure 8. Magnesium and calcium concentration

It would have been interesting to continue these testswith water containing bicarbonate and more chloride, butit was decided to use a different approach in order to studythe influence of bicarbonate and carbonate ions on corro-sion rates.

Immersion Tests with Steel Specimens

Because almost all natural waters contain bicarbonate inat least a small concentration, it was felt that considerationshould be given to different proportions of bicarbonate andother anions in corrosion tests with controlled syntheticsolutions containing known concentrations of sodium salts,eliminating the possible added influence of bivalent metalions. Borgman 20 has indicated the relative corrosiveness ofsalts of numerous cations and anions, exclusive of bicar-bonate and carbonate, and largely in concentrations greaterthan that in natural waters.

The effect of carbonate ions as an inhibitor of corro-sion was previously demonstrated by Evans 21 i n 1 9 2 7 .Mears and Evans 22 in 1935 described in detail the inhibitingeffect of potassium carbonate on solutions containing po-tassium chloride. These data, however, concerned strips ofsteel partially immersed in solutions of known concentra-tions, and provided no information on the pH of the re-sultant mixtures of carbonate and chloride salts. In otherwords, although the potassium carbonate concentrationvaried, the pH was not held constant but was different forthe various proportions. Therefore, pH as well as the car-bonate ion variables affected the results.

Apparatus of the standard type 23 for total-immersiontests of nonferrous metals was constructed for the WaterSurvey studies and is shown in figure 11. The 1½×3 inchtriplicate specimens of 0.01-inch ‘black plate’ steel (freefrom mill scale) that were used were reported to have thefollowing composition (by percentage): C, 0.07; Mn, 0.30-0.45; P, 0.015 maximum; S, 0.050 maximum; and Si, 0.010maximum.

The specimens were degreased in carbon tetrachloride;placed in a 5 percent solution of HCl and HNO3 for 2 min-utes; placed in concentrated HCl for 1 minute; rinsed in

14

2.0

Figure 9. pH and concentrations at cathode

acetone; and dried and weighed 48 hours before use. Theedges were coated with paraffin, and a scratch was made onboth sides of the specimen just prior to immersion in 18liters of water for a 3-day test at a flow velocity of 0.085fps at room temperature. The ratio of surface area of thespecimens to volume of water was 3 square centimeters perliter. The specimens were suspended from a bar that wasmechanically rotated in a vertical plane to move the sub-merged specimens in a similar vertical plane and produce anequal velocity on all points of the surfaces. A low velocitywas deliberately chosen in order to simulate the conditionsusually existing in as much as 25 to 50 percent of anymunicipal distribution system, including service lines.

Various proportions of sodium bicarbonate and chloride,sodium bicarbonate and sulfate, and sodium bicarbonateand nitrate were used at pH 7 and 9. The water was con-tinuously aerated, and carbon dioxide was continually bub-bled through the water to adjust and control pH. Theresults are shown in figure 12.

It was noted that, after the proportion of sodiumchloride or sulfate reached a given value, a corrosion rateof from 100 to 125 milligrams per square decimeter perday (mdd) was not influenced by further addition of thesechemicals. Also, the rate was of the same order of magni-tude whether chloride or sulfate was used. The corrosionrates with these proportions were therefore assumed to begoverned strictly by the dissolved oxygen content of thewater. In other words, if 15 or 20 mg/l dissolved oxygenwas present, the corrosion rate in this range may have beenhigher, particularly with increasing proportions of sodiumchloride. It may be worthy to note that the highest ratioof chloride to alkalinity that did not corrode was about onehalf as much as the sulfate to alkalinity ratio.

It was repeatedly found that the corrosion rate waszero when a particular minimum of alkalinity was presentin each test. It was also noted that an intermediate rangeof corrosion existed in which the rate was unpredictable Figure 10. Concentration in compartments

15

Figure 11. Total-immersion test setup

under the experimental conditions. For example, of threespecimens in a solution in this range, one may have corrodedat a rate of 10 mdd, and the other two may have corroded

at a rate of 90 mdd in the same solution. In this range ofwater quality, the corrosion rate may have been inhibitedor intensified, depending upon the sensitivity and discon-tinuities of the surface of the specimen.

It was significant that, for any specific chloride or sul-fate concentration, corrosion rates might be considerablygreater for solutions of low total mineral content than forthose of high total mineral content with appreciable alka-linity, a finding that is contrary to the usual predictionswhich forecast higher corrosion rates for higher mineralcontent.

Also, for some chemical compositions, corrosion ratesappeared to be greater at pH 9 than at pH 7; however, forothers, the rates were unchanged. This also is contrary tothe normal predictions on the corrosive tendency of water.

The relatively lower corrosion rates experienced withnitrates were surprising. Although a water that containsonly bicarbonate and nitrate is a rarity, it should be ofinterest to make a further study on the effect of small con-centrations of nitrate on corrosion rates in water containingvarious mixtures of chloride and bicarbonate.

Several spot tests with a 9-day immersion period yieldedresults similar to those from the 3-day data.

These data are specifically limited to dissolved solidsconcentrations between 200 and 1200 mg/l, under the flowvelocity and temperature conditions indicated.

Figure 12. Results of total-immersion tests

16

Figure 13. Effect of chloride-bicarbonate salts of sodium ratio on corrosion of mild steel

Figure 12, however, also shows the corrosion rates ex-perienced in a test series in which the combined sodiumchloride and bicarbonate concentrations ranged from 60 to250 mg/l. Again, it was noted that the bicarbonate exertedan inhibitive effect. In one group of tests with University ofIllinois tap water at pH 7 (-0.4 saturation index), no cor-rosion was noted until 60 mg/l NaCl was present.

Extreme caution is needed in interpreting these data orapplying the conclusions to other conditions. Considera-tion must be given to the fact that, at the low velocitiesemployed in these studies, the electrical migration of ionsunder the corrosion cell potentials plays a more importantpart in the process than the relatively slow diffusion rate ofthe dissolved oxygen. At a higher velocity, it might be ex-pected that oxygen diffusion rates would be the more im-portant factor. Also, the relatively high mineral contentminimizes the effect of pH because the hydrogen andhydroxyl ion concentrations are relatively low.

One severe criticism of these data is that no attempt wasmade to distinguish between general corrosion and pitting.Where pitting occurs, the rate of penetration may be quitehigh, although the corrosion per square decimeter of thetotal surface may be no greater than in areas where generalcorrosion is experienced. However, Mears and Evans22

have shown that pitting is less likely to occur where little

or no anodic inhibitor is present.Further immersion studies with mild steel confirmed the

conclusion that corrosion is accelerated by increasing pro-portions of chloride to bicarbonate. During this study 19

considerable time was spent in evaluating various methodsof preparing specimens, and it was found that very minorstresses imposed on the steel by lack of care in handlingpromoted excessive variability in results. Subsequent 3-daytests established a better corrosive-inhibitive relationshipfor various equivalent ratios of chloride-bicarbonate saltsof sodium at pH 7, as shown in figure 13. The curves indi-cate the upper and lower boundaries of the observed cor-rosion rates. This figure would appear to indicate thateven small proportions of chloride to alkalinity cause somecorrosion.

From these data it appears that increasingly high rates ofcorrosion occur as the chloride to bicarbonate ratio in-creases, particularly above a value of 0.3. This proportionapproximates the proportion of chloride plus sulfate to bi-carbonate in Great Lakes waters and classifies24 such waterin a range which is sensitive to additional corrosives andprobably to inhibitors.

C. H. Spaulding, in a preliminary unpublished study ofthis problem, suggested that free chlorine appeared to beresponsible for excessive tuberculation. This and other work

17

Figure 14. Effect of free chlorine on corrosion of mild steel

on the problem25 prompted the undertaking of a numberof tests to obtain a relative evaluation of this factor. Thesetests indicate (figure 14) that free chlorine in concentra-tions above 0.4 mg/l corrodes steel at room temperature inaerated water of about 120 to 135 mg/l alkalinity andabout 30 mg/l sodium chloride, at pH 7 and 8, at lowvelocities.24 In figure 15, these results have been superim-posed on other data for corrosion rates with no chlorineadded to indicate the relative corrosion rates at the particu-

lar chloride-bicarbonate ratio. Figure 15 also shows the re-sults of tests of 3 to 6 days duration using chloramine inconcentrations of 0.4 to 3.6 mg/l. These data suggest thatcorrosion is inhibited by conversion to chloramine.

It is probable that similar relative results would not havebeen obtained at other proportions of chloride to bicar-bonate. If 5 mg/l sodium chloride had been used ratherthan 30 mg/l, a greater proportion of inhibitor would havebeen present and more than 0.4 mg/l chlorine may havebeen necessary for comparable corrosion rates. If 80 mg/lsodium chloride had been used rather than 30 mg/l, theadditional corrosion induced by 0.4 to 1.0 mg/l chlorinemay have been negligible. No additional studies have beenmade on other corrosives or inhibitors.

Tests have also indicated corrosion rates to be higher atpH 8.5 and 9.0 than at 7.0, 7.5, and 8.0. A twofold in-crease in corrosion rate resulted when the velocity was in-creased from 0.14 to 0.89 fps at pH 8.5 and 9.0 with 1000mg/l alkalinity and 120 mg/l sodium chloride as shown infigure 16.

Minimum corrosion was noted at pH 7.0, 7.5, and 8.0at both rates. Although high pH is normally considered aninhibitor, this apparent contradiction indicates a need torecognize alkalinity as a natural inhibitor in water. In theabsence of calcium the high pH intensified incipient pitting.

Figure 15. Relative corrosion rates at particular chloride-bicarbonate ratios with and without chlorine

18

Figure 16. Effect on corrosion rates of increased pHand change in velocity

As previously mentioned, a new method for measuringcorrosion rates on submerged specimens was developed in1956. 18 For the first time, rates of corrosion could bedetermined in a very short time without destroying thespecimen. Thus, a single specimen would suffice forperiodic measurements over an extended period of time.

The method consists of applying current to the specimenin micro amounts and measuring its change in potential

Figure 17. Empirical relationship between initial slopeof polarization curve (resistance) and corrosion rate

determined by weight loss

against a reference electrode. The change in potential perunit of applied current density (∆E/cd) was found to cor-relate quite well with weight loss measurements convertedto milligrams per square decimeter per day as shown infigure 17. It is, in effect, a measure of film resistance on themetal surface.

This method was used to establish the data on the effectof pH on corrosion of steel in water of a specific quality(see figure 18). The curves indicate results after: A, 16days; B, 12 days; C, 8 days; D, 4 days; and E, 2 days. Testswere made on an air-saturated solution with a NaHCO3 con-tent of 2.5 me/l, NaCl content of 0.5 me/l, temperature of19 to 28°C, and a velocity of 0.14 fps. The rates of cor-rosion were confirmed by total weight loss of the specimensat the end of the test.

Figure 19 shows the results of a more recent test26 witha water having a much higher chloride content than thatshown in figure 18. The general trend of the data in bothof these figures is similar; however, the corrosion rates inwater having high chloride content are considerably higherthan those in water of a low chloride content. Both of thesetests were conducted at low velocities.

It should also be noted that the alkalinity was constantat all values of pH. The Whitman, Russell, and Altieridata 3 were obtained with different concentrations of alka-linity and chloride at each pH (see page 7).

Figure 18. Effect of pH on corrosion rateat chloride-alkalinity ratio of 0.4

19

sity) equals inches per year. The ipy measurement mayalso be made for the depths of pits on the surface. Thesecan be related by considering no corrosion over part of thesurface as the result of being cathodically protected by theremaining surface (the anode) which is corroding.16

When essentially no protection exists, there is a generaluniform corrosion rate (figure 21). If part of the surface isnot corroding, the remainder of the surface is corroding ata proportionally greater rate. Deep pitting or local pene-tration can take place when the surface is predominately ca-thodic, unless the corrosion products become so dense thatthe overall rate is reduced. Local points of tuberculationare associated with local points of pitting.

Mild Steel

These studies, in part, contribute to the following classi-fication of waters which considers the mineral content, dis-solved oxygen, and pH (acidity) relative to the effect on thecorrosion of mild steel.

No Minerals – Dissolved Oxygen Present. In the ab-sence of minerals, increasing pH decreases corrosion ratesby water containing dissolved oxygen. However, if the pHis near, but not above, that required for complete protec-tion, pitting occurs, which rapidly decreases the useful lifeof mild steel. Crevices at joints and welds and corners wherecirculation is poor, which do not permit oxygen to be main-tained at the surface, are subject to localized corrosiveattack.

Pitting takes place at local unprotected points of cor-rosion where the corrosion products prevent the diffusionof oxygen to the metal surface, thereby permitting dif-ferences in oxygen concentration at the metal surface. Sim-ilar pitting can occur under deposits of debris. Also, dif-ferences in oxygen concentration, as at the water line of sur-faces exposed partly to air and partly to water, will causepitting. The rate of corrosion increases at higher tempera-tures.

No Minerals – Dissolved Oxygen Absent. In the absenceof dissolved oxygen, the pH of ‘high purity’ water confined

Figure 19. Effect of pH on corrosion rate with Cl/alk ratio of 1.0

Figure 20 shows the type of corrosion products adheringto the steel specimens used to obtain the data in figure 19.These pictures were taken at the end of the test (35 days),and the corrosion rates shown were final measured rates.It is interesting to note the increasing degree of tubercula-tion as the pH increases.

Significance of Corrosion Rate Measurement

When a corrosion rate is measured, it is usually reportedin milligrams per square decimeter per day (mdd) or interms of inches of penetration per year (ipy). The first is aweight measurement per unit surface; the second is theaverage depth measurement on the surface, calculated fromthe weight measurement. Therefore, mdd x (0.00144/den-

20

Figure 20. Corrosion products on steel specimens

Figure 21. Relative corrosion rates with partial protection

21

in mild steel containers adjusts itself to about 8.4 and cor- However, such increasingly high concentrations arerosion becomes negligible. However, all natural waters are responsible for an additional tendency to deposit ob-mineralized to some degree. jectionable quantities of scale at temperatures above

Noncarbonate Minerals – Dissolved Oxygen Present. In that at which saturation stability is established.the absence of carbonate minerals, increasing concentrations 3) The protective action is enhanced by movement ofof other common minerals, such as chloride and sulfate the water and decreases at near-stagnant conditions.salts, increase the conductivity and therefore the corrosion 4) The protective action may be nullified at higher tem-rate at all pH values below the pitting range of pH, and in- peratures when the pH is high enough to deposit non-crease pitting when the pH is just below that required for protective magnesium hydroxide.uniform protection in the presence of dissolved oxygen. In- 5) Pitting and tuberculation will occur in the presence ofcreasing temperature accelerates both general corrosion and dissolved oxygen if the stability and velocity are near,pitting. but still below, that required for complete protection.

Carbonate Minerals (No Calcium) – Dissolved Oxygen 6) The protective action is decreased by increasing pro-Present. Carbonate minerals, indicated by the alkalinity portions of chloride and sulfate salts above a ratio ofdetermination for bicarbonates, inhibit corrosion. They act about 0.1 or 0.2 to 1 with respect to alkalinity. Thiscontrary to the accelerating salts of chloride and sulfate in limitation becomes less significant in the absence ofwaters containing dissolved oxygen. In the absence of cal- dissolved oxygen.cium, this inhibition is maximum at pH 6.5 to 7.0 in con-centrations five to ten times above the chloride and sulfatesalt concentration. Inhibition also is greater at increasing pH

Cast Iron

above 9. At concentrations decreasing five or ten times be- Corrosion of cast iron is unlike that of steel. It appearslow the chloride and sulfate salts, corrosion rates increase. that a great difference in susceptibility of cast iron to cor-Since nearly all natural domestic waters or carbonate min- rosion exists between ‘as cast’ or annealed cast iron, anderals contain alkalinity, and in addition usually contain machined specimens with no protective barrier.chloride and sulfate salts, this too is a criterion in classifi- Modern production of cast iron produces an outside sur-cation. face of iron oxide, silicates, and alumino silicates which

Minerals – Dissolved Oxygen Absent. In the absence of serves as an excellent barrier to atmospheric corrosion, anddissolved oxygen, the types of mineralization are less impor- usually to water. When this surface is ground to make ittant with respect to useful life of mild steel. This is indi- smooth or machined for fabrication, a surface containingcated by the following examples: graphite in a matrix of ferrite or pearlite is exposed. The

1) Plumbing experience indicates that properly designed graphite, which occurs at about 0.04 millimeter intervals,hot water heating systems can be made of steel, pro- serves as cathodic points to stimulate an initial galvanic cor-vided that no water (domestic) is lost from the sys- rosion of the adjacent iron.tem; avoiding the addition of make-up water con- Grinding of Cast Iron Pipe Interiors. It was learnedtaining dissolved oxygen also prevents loss of corro- through plant inspection and correspondence that the inte-sion products that inhibit the corrosion. rior of all centrifugally cast pipe is partially ground for the

2) The useful life of steel used for feedwater heaters, purpose of reducing roughness prior to the application ofboilers, and piping for power plants may be extended coal tar coating or cement lining.26 In this process, mostconsiderably by maintaining an oxygen-free boiler cast iron pipe interiors are ground to the extent that 10 tofeedwater. 50 percent of the surface consists of graphite in the ferrite

Calcium Salts – Dissolved Oxygen Present. From the or pearlite matrix.standpoint of corrosivity, stability as indicated by satura- Because most cast iron pipe in service is tar coated, ittion with calcium carbonate is the most widely accepted appears that tuberculation in such pipe results from slightcriterion in classification. Uncoated mild steel or machined but significant corrosive action of partially inhibited watercast iron requires a very significant supersaturation to form on holidays, or minute holes, that occur in the tar coatinga visible deposit. This criterion has specific limitations: covering the machined areas.

1) A minimum alkalinity of 50 to 100 mg/l (calculated To determine the composition on the interior ‘skin’ ofas CaCO3) and a minimum calcium of about 50 mg/l the unground pipe surface responsible for the remarkable(as CaCO ) must be present at normal temperatures corrosion resistance, a specimen was submitted to Professor3

(32 to 160°F) for an extended life for the metal. W. H. Bruckner of the University of Illinois Department ofSuch water will require a pH much higher than the Metallurgy. Figure 22 shows a microphotograph (magnifiedcalculated pH of saturation. 50x) of an unetched mount.

2) The greater the concentrations of calcium and alka- When the hard ‘skin’ on the internal surface of the pipelinity, the greater the protective action of the water. was scraped off, the black phase between the metal surface

22

Figure 22. Microphotograph of an unetched mount of cast iron

The external gray phase on the exposed surface, adja-cent primarily to the magnetite, showed two or more com-ponents under polarized light: long needle-like pyroxenesof monoclinic structure as CaSiO3 or FeSiO

3(Mg or Mn);

and polyhedron garnet of cubic structure, such as Ca3Al

2(SiO4 )2 or Fe3

Al2

(SiO4 )2 .

(with black graphite particles dispersed in the matrix) andthe gray phase was found to be ferromagnetic (Fe3O4 )with a cubic structure.

This silicate skin, which is highly insoluble, appears toprovide an effective barrier to corrosion by water. It isvery brittle and therefore may not be a completely effec-tive barrier when damaged.

Immersion Tests with Cast Iron Specimens. Laboratorytests were performed with cast iron in a manner similar tothe tests for steel. The specimens were totally immersed inthe water, saturated with air at room temperature, and withcarbon dioxide used for controlling pH. The apparatus, asused with the steel immersion tests, provided a constantvelocity of the specimen in a vertical circular path. Castiron specimens were machined and abraded with No. 120grit paper and degreased prior to use. At the end of thetests, the specimens were cleaned electrolytically for 5 min-utes in 10 percent ammonium citrate prior to the deter-mination of weight loss, in terms of milligrams per squaredecimeter per day.

Calcium Effects. With machined cast iron specimens inwater containing calcium, an interesting overall average ofthe results is indicated in figure 23.27 This broad generali-zation indicates that irrespective of the saturation index, ofvelocities from 0.08 to 0.85 fps, of minor variations in thechloride to alkalinity ratio, and of the presence or absenceof chlorine, chloramine, or silica, the concentration of cal-cium in the presence of 125 mg/l alkalinity is a primary fac-tor in the inhibition of corrosion of machined cast iron. Italso shows that a certain length of time is required for ef-fectiveness to become apparent. In this generalization rep-

It is significant to note in both figures 24 and 25, thetuberculation was greater at the –0.3 saturation index thanat either the higher or lower levels, and the corrosion ratecontinued at a high rate for a longer period of time. Thisis more striking at the low velocity in the figure 24 experi-ment than in figure 25.

Figures 25a and b show almost identical features. Figure25b includes clear evidence of the effect of adding 56 mg/lsodium chloride to the –0.3 saturation specimens.

Parenthetically, the free- and combined-chlorine levelswere maintained in these tests without substantial increasein chloride content by periodically replacing portions of thewater with fresh water and adjusting the mineral composi-tion.

With free chlorine added (figures 24a and b), it had beenobserved that round knobs of 3/16-inch diameter protrudedas much as 1/8 inch. Similar tubercles, but smaller in sizeand number, were noted in the –0.3 saturation index en-vironment when free chlorine was absent. After the speci-mens were cleaned, graphitization was observed where thetubercles had been, clearly indicating localized corrosion.

Figures 25a and b provide similar data at 0.34 fps, butwith combined chlorine present rather than free chlorine atpoint B. The more rapid reduction and the lower final cor-rosion rates testify to the effectiveness of a greater velocityto improve the diffusion rate of calcium as the inhibitor.Reducing the saturation index gave no change in the corro-sion rate because of substantial protection previouslydeveloped.

At point A, 56 mg/l sodium chloride (1.0 me/l) was add-ed, and the saturation index was reduced to –0.1 and thepH to 7.5. The effect of sodium chloride is apparent. Theaddition of 5 mg/l silica at this point had no apparent effecton the corrosion rate, as can be seen in figure 24b.

The specimens in –0.3 saturation index water leveled offat a higher final corrosion rate than either of the other two.It should be indicated that the specimens at this satura-tion index showed small dark reddish spots or growths sur-rounded by an overall whitish red deposit prior to the addi-tion of chlorine, and continued to form tubercles. To alesser extent, similar growths developed on the specimensin +0.2 saturation index water.

For the following specific modifications in water quality,the data are represented as smooth lines from the averagerates determined five times per week for triplicate speci-mens. Figures 24a and b (parallel tests) show the reductionof corrosion rates with time for three degrees of saturationwith respect to calcium carbonate at a low velocity of0.085 fps. The amount of calcium and alkalinity was 125mg/l (as CaCO ). The decrease in rate of corrosion occurredmost rapidly for the water with a saturation index of +0.2and least rapidly for the –0.8 saturation index water.

resenting all tests, there are broad variations, as much as ten-fold in some cases, for each level of calcium.

3

23

Figure 23. Summary of calcium effects on corrosion

3

The influence of the chloride-alkalinity ratio is shown infigure 26 by comparing the top and bottom curves; theintermediate ratio of 0.3 is also indicated by the two mid-dle curves. The relative influence of free and combined

Although the calcium concentration was lower than infigures 24 and 25, the rate of corrosion after 1 day was lessthan that when 125 mg/l calcium hardness was present. Therate of approach to equilibrium was lower and the final cor-rosion rate was higher.

Chloride to Alkalinity Ratio Effects. To consider fur-ther the relative influence of the calcium concentration,saturation index, chloride-alkalinity ratio, and chlorine, theresults of four test conditions are indicated in figure 26.27

All tests were made with a water velocity of 0.85 fps. Cal-cium concentration (as CaCO ) was 31 mg/l; alkalinity (asCaCO3 ) was 25 mg/l. The saturation index was –0.25, andthe pH (except as indicated) was 8.0. These machined castiron specimens showed tuberculation in all cases, with orwithout chlorine or chloramine, and were similar in appear-ance to the –0.3 saturation index specimens as indicated infigures 24 and 25 with 125 mg/l calcium hardness.

chlorine appears to be consistent with the observationsin figures 24 and 25, but of only minor significance withthis already more corrosive water.

Of particular significance is the rise in corrosion rate inone of the solutions at the 84th day, when the pH was per-mitted to drop to 6.7. On detailed examination of the data,it often appeared that an inadvertent drop in pH of 0.1unit increased the corrosion rate by about 8 mdd.

From an applied standpoint, it is important to recognizethat liquid chlorine is an acid, and converts alkalinity to car-bon dioxide.16 This results in a reduced pH and conse-quently a reduction of the saturation index. Therefore astable or protective water may be converted to one withcorrosive tendencies. Just as alum (which is an acid whendissolved) requires neutralization by lime, caustic soda, orsoda ash, so does liquid chlorine. For each mg/l chlorineadded, 1.13 mg/l caustic soda is required for the neutraliza-tion. Of course, sodium hypochlorite can generally be sub-stituted for liquid chlorine in small plants, with no need forneutralization.

pH Effects. One study was conducted to show the ef-

24

Figure 24. Tests made at 0.085 feet per second water velocity

fect of pH on the corrosion rates of cast iron at a flow rateof 0.1 fps for a zero hardness water having an alkalinity of125 mg/l and a chlorine-alkalinity (Cl/alk) ratio of 0.2. 28

The results shown in figure 27 are generally similar to thoseobtained for steel specimens (see figure 19) exposed tosimilar water with no hardness and a chloride-alkalinityratio of 0.15 at a velocity of 0.14 fps, in that the lowestaverage corrosion rate occurs at pH 7.0.

By comparison with the results for cast iron specimensillustrated in figure 26, this study with no calcium at pH8.0 showed a corrosion rate of about 65 mdd at 100 days.Severe local pitting was noted adjacent to the paraffin coat-ing over the edges.

Tuberculation on the machined cast iron surface in-creased with increasing pH in the range 7.5 to 9.0. This isin contrast to the general corrosion at pH 7 and lower. Thethick corrosion products at pH 7 appear to be effectivelyresistant to the diffusion of dissolved oxygen, and therebyreduce the corrosion rate.

A similar study was made to determine the effect of theCl/alk ratio on the corrosion of machined cast iron in waterswith an alkalinity of 120 mg/l at calcium concentrations of

Figure 25. Tests made at 0.34 feet per second water velocity

30 and 85 mg/l. In this study, the pH was regulated to28

maintain a zero saturation index and the velocity was 0.8fps. The corrosion rates are shown in figure 28. It can beseen that the relationship of the Cl/alk ratio to the corro-

Figure 28. Effects of Cl/alk ratio and chlorine

25

Figure 27. Relationship of corrosion rate to pH

sion rate is minimized within 30 days when the water isstabilized with respect to CaCO3 . The beneficial effect ofhigher calcium hardness is shown even though the pH islower.

In another series, tests26 were designed to determine therelative influence of sulfate and chloride ions on tubercula-tion. No significant differences from the foregoing studieswere noted with regard to corrosion rate or tuberculation inconcentrations up to 30 mg/l sulfate and 3 to 25 mg/l chlo-ride with water of an 85 mg/l calcium hardness, a 120 mg/lalkalinity, and a 0.29 (Cl + SO 4 )/alkalinity ratio at satura-tion indexes of -0.8, -0.3, and 0.0.

Figure 28. Decrease in average corrosion rate with timeat 0 saturation index

26

When the pH was changed during the test from a 0.0saturation index to a lower level, the corrosion rate in-creased. When the pH was changed from a negative satura-tion index to a 0.0 index, the corrosion rate decreasedstrikingly.

Relative Corrosion Rates. Sixty-seven experiments wereconducted to determine the relative corrosion rates at roomtemperature of uncoated cast iron specimens (without natu-ral scale protection) in oxygen-saturated, synthetic watersof different mineral composition which are not unusual forpublic water supplies.16 As in previous studies, the ex-periments were made at constant velocity with pH adjustedby the controlled addition of CO2 . The instantaneous cor-rosion rate measurements were made by momentary polari-zation of the specimens twice a week. Tests were con-ducted long enough to establish a constant or equilibriumrate. This usually required an exposure of 3 to 4 months ormore.

A summary of equilibrium rates is shown in figure 29.The velocity was 0.1 to 1.0 fps, and the temperature was20°C. The initial rates were usually 100 to 150 mdd, andgradually approached the rates indicated in figure 29. Themineral variables were alkalinity, calcium, chloride, and so-dium with the pH adjusted to the various levels indicated.The apparent general characteristic of these data is a maxi-mum rate of corrosion at a particular pH level, with reducedrates at other pH levels (Ca and alk are expressed as CaCO

3 ).

The lower minima at about pH 6 to 7 represent evenlydistributed general corrosion with a relatively thick layer ofcorrosion product which provides a protective barrier tocorrosion. This product occasionally flakes off with a tem-porary increase in rate, which decreases as the thicknessagain increases. No tuberculation is evident. At somewhathigher pH levels the corrosion rate increases and tubercula-tion becomes apparent, particularly at the maximum corro-sion rate and before the minimum rate at the upper pHlevel. In this range, cleaned specimens show areas of dis-tinct graphitization similar to pitting attack under the pointsof tuberculation. The upper pH minima occur at or abovethe pH of saturation for calcium carbonate. At the upperpH minimum, the deposit is slightly rough but hard andeither black or coated with a calcium deposit.

Similar characteristic maxima with respect to pH wereobserved by Uusitalo and Heinaneu29 in Finland on steel,after the data had been replotted with respect to pH (figure30).

Supplementing these studies, Stumm30 conducted ex-periments to relate corrosion rates to the amount of cal-cium carbonate deposited on cast iron. An apparent con-flict was evident by some experiments indicating a heavydeposit of calcium carbonate at pH 9.5 with an accompany-ing high corrosion rate, under conditions of high saturationindex. Other experiments indicated very low corrosion rateswith a very small deposition of calcium carbonate under

Figure 29. Summary of equilibrium rates

Figure 30. Corrosion of steel specimens

27

conditions of a slightly negative saturation index. This wasattributed to the pH effect.

The high pH water had a calcium content of 55 mg/l (asCaCO3) and an alkalinity of 50 mg/l (as CaCO3 ) which pro-vided only a minimum buffer capacity at pH 8.5. The lowpH water (pH 7.1) had calcium contents of 190 and 250mg/l and an alkalinity of 250 mg/l with about a 50-foldgreater buffer capacity.

It was recognized that calcium carbonate is basically acathodic inhibitor and deposits at the cathodic areas of thesurface. Microscopic examination of the specimens in thehigh pH water showed the protected cathode areas to bequite large, and the anodic areas also quite large, as dis-tinguished from observations on the specimens in the lowpH waters where the anodic areas were numerous and verysmall, as in general corrosion. This latter condition permit-ted clogging of these small anodic areas with the corrosionproduct and calcium carbonate, as indicated by significantennoblement of the corrosion potential after 20 days of ex-posure, whereas the high pH water was unable to block thelarge active anodic areas, and little ennoblement occurredduring the exposure.

Thus, the combination of high alkalinity and calciumwith the inherent low pH, can prevent corrosion more easilythan low alkalinity and calcium, even with a high pH and apositive saturation index.

Summary

The finding that calcium in the presence of alkalinity,regardless of pH or saturation index, is an effective inhibitorto corrosion is not inconsistent with theory and experience.The basic corrosion reaction at the cathodic points is con-ducive to the formation of calcium carbonate at the point

of corrosion even though the water environment is belowsaturation with respect to calcium carbonate. The rate atwhich such deposition may take place is increasingly de-creased by higher concentrations of calcium and/or alkalini-ty. The rate is further improved by pH adjustment up to orabove the saturation pH for calcium carbonate, and underthe higher pH levels the rate is also increasingly improved byhigher velocities.

The finding that tubercles form under conditions below,but near, complete protection is consistent with the well-known action of all cathodic inhibitors to induce pittingcorrosion, rather than general corrosion, when present in aconcentration that is insufficient to produce complete pro-tection. This relation is confirmed by the observation ofpitting corrosion or graphitization under tubercles on thespecimens. The fact that tubercles do form apparently isdue to the deposition of a supporting structure of calciumcarbonate with or in the normal corrosion product of ferrichydroxide in aerated waters.

If the saturation index is not sufficiently close to zero orthe concentrations of calcium and/or alkalinity are too low,a supporting structure may not be developed, although ap-preciable protection may be afforded by calcium carbonateimbedded in the rust deposit from generalized corrosion. Itwas observed on specimens in waters of -0.8 saturation in-dex with high calcium at pH 6.8 (see figures 24 and 25) thatthe specimens were coated with a whitish red deposit of suf-ficient thickness (about 1/16 inch) and rigidity to continueto adhere even though this deposit was cracked for distancesof 0.5 inch.27

There were no pits under this deposit and itserved as an effective barrier to the diffusion of oxygen andother corrosive agents.26

The finding that chloride accelerates corrosion of castiron is consistent with the observations on steel specimens.

CORROSlON PROBLEMS IN HOUSEHOLD AND INSTITUTIONAL FAClLlTlES

In contrast to central public water supply distributionsystems, private homes, apartments, business and commercialbuildings, and institutional buildings and grounds use waterfor many purposes, with multiple accompanying problems.Experience has shown that materials of construction are bas-ic to corrosion control, and a variety of materials are avail-able for selection as well as coatings for protection. Dif-ferences in water supply qualities may be more or less ag-gressive to different metals and may dictate the primarychoice. A more expensive selection of piping and appurte-nances can avoid costly chemical treatment. Minor modifi-cations in water quality in some instances may be able tocorrect a severe problem. However, water treatment alonecannot be expected to control corrosion under adverse con-

28

ditions of poor choice of construction materials for specificuses, of inadquate design, and of excessive temperatures.

Materials of ConstructionAlthough materials of construction are necessarily selec-

ted on the basis of their physical and mechanical properties,corrosion resistance is an important consideration for main-tenance and the useful life of the installation. Lack of ex-perience and knowledge to anticipate problems of leaksand failures from corrosion, or the contamination of thewater with toxic metals or esthetically unacceptable prod-ucts of corrosion, can be costly.

Cost considerations for the selection of materials should

repair, the life expectancy of the installation, and watertreatment requirements if necessary.

Steel, iron, galvanized steel, copper, and aluminum arereadily available at reasonable cost. Brasses are generallyused for valves and fittings, and stainless steels are availablefor special needs if necessary. Plastic pipe and fittings arecompetitive and useful, where the high coefficient of ex-pansion with temperature is not prohibitive.

Mild Steel. The process of manufacturing and the com-position do not significantly affect the corrosion rate ofmild steel in domestic waters. Copper bearing steel withproven effectiveness in air is no better in water than mildsteel without copper. The presence of mill scale on the sur-face accelerates pitting by galvanic action.

Dissolved minerals with increasing chloride and sulfateconcentrations increase the conductivity of water and maythereby increase the corrosion rate. With increasing propor-tions of alkalinity in the form of bicarbonate to the chlo-ride and sulfate, and increasing proportions of calcium hard-ness to sodium and magnesium, the corrosion rate can beinhibited (see page 7 for stability). For any given waterwith insufficient calcium hardness and alkalinity, the cor-rosion rate is proportional to the concentration of dissolvedoxygen and, generally, to the rate of flow. At low flowrates where the corrosion product (rust) remains on the sur-face, the rate of diffusion of oxygen to the surface is hin-dered and the corrosion rate is reduced until the depositflakes off.

Elevated steel tanks should be painted periodically in ac-cordance with AWWA Standards (D102-64), and shouldhave properly installed and controlled cathodic protectionsystems.

Zinc and Galvanized Steel. Zinc like aluminum andlead is an amphoteric metal and corrodes in alkaline as wellas acid solutions, In the absence of carbon dioxide or thecarbonate species, it has minimum solubility in a relativelynarrow pH range near 9.3.31 In distilled water corrosion in-creases with rise in temperature from 120°F to a maximumat about 160°F where the corrosion product is not adherentand then decreases above 160 to 212°F as the solubility ofoxygen decreases.32

In distilled water at 85°F, over a range of controlled pHlevels from 2.6 to 14.4 with HCl and NaOH, the corrosionrate of zinc was less than 0.05 inches per year between pH6 and 13.33

Increasingly rapid corrosion was recorded asthe pH deviated below 6 and above 13.

In natural waters corrosion of zinc is dependent uponthe pH, alaklinity, and type of dissolved minerals. Withadequate carbonate hardness (alkalinity and calcium), pro-tection by basic zinc carbonate is often effective at or nearthe pH of saturation for calcium carbonate.

Galvanized steel relies on a zinc coating for protectionof the steel from corrosion. The zinc protects the steel by

excluding water contact with the steel and also by galvanicprotection, where the anodic zinc corrodes more easily thancathodic steel. Zinc is sacrificed for the protection of thesteel at holidays and other exposures. If on the other handthe zinc corrosion products completely protect the zinc, itssacrificial property is negligible and pitting at the steel ex-posure can take place, unless the steel was galvanically pro-tected by calcium carbonate at early exposure.

In waters containing calcium bicarbonate, anodic dis-solution of zinc develops a film of basic zinc carbonate atanodic areas and a film of calcium carbonate at exposediron areas. Such deposits may be hindered by relativelyhigh concentrations of sulfates and chlorides, and by insuf-ficient concentrations of calcium and alkalinity. Large areasof exposed steel at joints and threads are particularly af-fected by these conditions. Traces (0.1 mg/l) of solublecopper can also be detrimental to zinc coatings, and gal-vanic corrosion of zinc is accelerated by contact with cop-per, or copper bearing metals.

Aluminum. The corrosion resistance of aluminum is dueto the protection developed by an inert oxide film. Thisfilm can be destroyed by acidic conditions (pH ≤ 7). Thepresence of copper, and high chloride concentrations, cangenerate local corrosion in the form of pitting. Aluminumshould never be used in systems involving copper bearingmetals.

Copper. The corrosion resistance of copper is usuallydue to an oxide film, and the various grades of copper dif-fer little in resistance to corrosion. The loss of resistanceby destruction of the oxide film does not necessarily lead tosevere attack as with aluminum. Copper is subject to severeimpingement attack at high rates of flow (>4 fps) particu-larly with soft waters with appreciable bicarbonate alkalinityand carbon dioxide. The presence of dissolved oxygen insuch waters has also been implicated. The impingement canbe particularly severe with high turbulence at elbows andconstrictions to flow. The impingement takes the form ofpits, usually in a ‘horseshoe’ shape.

General corrosion of copper at low velocities is enhancedby insufficient calcium hardness and alkalinity, accompaniedby chloride and sulfate. Dissolved copper causes blue togreen stains on porcelain ware. (Copper should never beused with low pH carbonated water.) Hot water tempera-tures enhance the rate of corrosion by such waters. Pittingcorrosion can take place with hard waters where debris orparticles of cement or sand adheres to the surface and pre-vents access of dissolved oxygen to the metal surface.

Copper Alloys. There are a number of copper alloyswith zinc and nickel. The more common copper-zinc alloysfor household systems are yellow brass 67-33 copper-zinc,and red brass 85-15 copper-zinc. The flexibility of copperpipe is sacrificed for strength, and often for better corro-sion resistance. The yellow brass is subject to dezincifica-tion in soft waters with appreciable carbon dioxide, and

29

always include cost of maintenance, ease of fabrication and

with water of high pH (>9.0 to 9.6), particularly with highchloride and/or at high temperatures. Red brass is preferablefor piping and valves with such waters.

The copper-nickel alloys appear to be superior for dif-ficult conditions. The 90-10 and the 70-30 copper-nickelalloys generally provide good service in hot water heatertemperatures. If 1 percent iron is included in the product,corrosion resistance to high velocities is improved.

Stainless Steel Alloys. There are a great number oftypes of stainless steels with different physical and mechan-ical properties, depending upon heat treatment and anneal-ability. The properties of these alloys often dictate choiceand often with no attention to corrosion resistance behavior.However, under certain conditions, their resistance to cor-rosion is far from good. Other selections are made withgreat care for corrosion behavior for acids or sea water ex-posure.

Chromium-iron alloys at 3 percent Cr show some im-proved corrosive resistance over steel, but for true ‘stainless’properties at least 12 percent Cr is required. However, evenat higher chromium percentages, there is still a risk withhigh chloride waters. The 11 percent chromium-iron alloytubing (Type 409) represents an example of poor corrosionresistance at the seam, particularly in waters with highchloride concentrations.

The corrosion resistance of true stainless steel is due to avery thin protective oxide on the metal surface. Mainte-nance of this film is dependent upon oxygen. Lack of oxy-gen can develop in crevices and under debris, with resultinglocalized corrosion. For this reason surfaces should bemaintained in a scrupulously clean condition. Type 304(18 percent Cr, 8 percent Ni) is a general all purpose stain-less steel that is used quite frequently in water treatmentplants and for well screens. Where pitting attack may beexperienced or anticipated, Type 316 (18 percent Cr, 8 per-cent Ni, 2 to 3 percent MO) offers greater resistance. Thepresence of chloride causes susceptibility to stress-corrosioncracking. Related factors in stress-corrosion cracking aretemperature, concentration of chloride, and degree of stress.For this reason stainless tubing has yet to be accepted forhousehold waters, although it is extensively used for foodpreparation, cutlery, table tops, sinks, and other householdfacilities.

For environments of excessive temperature, chloride, andstress, where stress-corrosion cracking may be anticipated,there are other more highly alloyed grades of stainless steelavailable.

Galvanic effects with copper or copper alloys are negligi-ble when coupled to 18-8 stainless steel.

Water Quality Considerations

Because of the limitation of treatments for householddistribution systems for drinking water quality, the first line

30

of defense in corrosion control must be the selection ofmaterials. Acceptable piping may generally be red brass orbronze, yellow brass, copper, 90-10 copper-nickel, or gal-vanized iron. The selection should be based on the mineralquality of the water and the need for and selection of treat-ment alternatives. The following assessments and recom-mendations can be made from experiences in state institu-tions in Illinois.

Cold Water Corrosion and Treatment

Surface Waters. Water from streams and lakes containdissolved oxygen, approaching saturation concentrations.Such waters are normally clarified, filtered, and chlorinated,and frequently (too often) are not adequately adjusted tothe proper pH for stability with respect to scale and corro-sion control.

For waters of high alkalinity and hardness, greater than150 mg/l, copper service lines may be preferred. However,galvanized steel may prove satisfactory, particularly if thepH is raised to slightly above the stability pH of saturation.If the chloride and sulfate level exceeds 150 mg/l, supple-mentary treatment with 8 mg/l liquid sodium silicate (asSiO2 ) can be beneficial.

For low alkalinity and hardness surface waters, less than150 mg/l (as CaCO3 ), copper is preferred. Galvanized ironmay be satisfactory if the pH is adjusted to the pH of cal-cium carbonate saturation, or higher. If the chloride andsulfate level exceeds 80 mg/l, sodium silicate formulationcan be beneficial.

A 2-year exposure of various pipe materials to waters ofdifferent qualities at 19 cities in the United States andCanada34 showed no corrosion of aluminum piping in limesoftened water at 7 cities. The greatest hazard with alumi-num, is the presence of copper or a low pH.

Groundwater. For groundwater with high hardness andalkalinity (>150 mg/l), with characteristically little or nodissolved oxygen, and with a combined chloride and sul-fate concentration of less than 150 mg/l, galvanized iron orcopper piping is satisfactory. For higher chloride and sul-fate levels, copper is the preferred material.

If such water is softened by ion-exchange, hard watershould be blended with the softened water to provide ahardness of 60 to 100 mg/l, and galvanized iron can be satis-factory with low chloride and sulfate, but copper is pre-ferred with higher levels. Supplementary treatment withsodium silicate (8 mg/l as SiO2 ) at pH 8.0 to 8.4 can be ben-ificial to copper as well as galvanized iron.

Hot Water Corrosion and Treatment

Corrosion of galvanized piping is more severe above140°F. Maintenance of hot water temperatures below amaximum of 140°F is an important factor in controlling

corrosion. If higher temperatures are required for dish-washing or laundry purposes, such systems should bedesigned for periodic replacement of common piping mate-rials, or should be constructed of more corrosion resistantmaterials such as red brass, 90-10 copper-nickel, stainlesssteel, etc.

In hot water (140°F) piping, ‘zero softened’ watercauses very serious corrosion of galvanized steel piping.Blending with hard water to provide 60 to 100 mg/lhardness has generally provided successful corrosion inhibi-tion. With highly mineralized water (Cl

– + SO 4

– – > 300

mg/l), liquid sodium silicate accompanied by a pH increasewith caustic soda from pH 7.9 to 8.4 is beneficial, if thealkalinity exceeds 150 mg/l.

Compatible materials should be installed in the samesystem; for example, cement-lined hot water heaters shouldbe installed with galvanized piping. Copper steam coils ordip tubes should never be used with galvanized hot waterheaters.

Hot water systems employing copper tubing should bedesigned for flow rates of less than 4 fps.

In rehabilitating old buildings and adding buildings to aninstitution already served by a central circulating hot watersystem with compatible materials, the same type of pipingshould be installed in the new buildings as in the old.

Sodium polyphosphate, which is ineffective as a corro-sion inhibitor under low velocity conditions, is also not con-sidered a generally effective inhibitor under other velocityconditions encountered in water distribution systems unlessconcentrations exceed 15 to 25 mg/l.

Treatment Evaluations. In an effort to evaluate mate-rials for institutional and domestic hot water distributionsystems for a number of different types of water qualities,and to investigate a variety of chemical treatment methods,controlled laboratory and field studies were made by theState Water Survey at several Illinois state institutionsfor the Department of the Air Force. 3 5 , 3 6 , 3 7 The selec-tions were based on the quality of water, the feasibilityfor altering the quality for controlled experiments, andspace requirements for the controlled studies.

Five types of water were included in the study with theuse of steel, copper, 90-10 copper-nickel, and galvanizedsteel. In addition to a control exposure in the unalteredwater, the following treatments were tested: liquid sodiumsilicate with or without pH adjustment, zinc silicate * withsulfamic acid, zinc polyphosphate, * * and zinc monobasicphosphate * * * with sulfamic acid.

* Zinc silicate formulation8 mg/l SiO2 (as liquid silicate) + 3 mg/l Zn+ +

(as ZnSO4 • H 2 O) + 1 mg/l sulfamic acid(as HNH 2 SO 3)

** Zinc polyphosphate formulationU. S. Patented formula by Calgon, Inc.

Zinc polyphosphate, composed of:

Phosphorus pentoxide (P2 O5) 56.5%

Zinc oxide (ZnO) 11.0%Sodium oxide (Na2 O) 32.5%

* * * Zinc monobasic phosphate formulationU. S. Patent 3669616 formula by Virginia Chemicals,

Inc.Zinc sulfate (ZnSO • H

4 2O ) 55.3%

Sulfamic acid (HNH2 SO 3 ) 20.0%Sodium monobasic phosphate

(NaH 2 PO ) 24.7%4For each of five types of water with no treatment, 90-10

copper-nickel was found to be preferable to copper, at 140and 180°F, and both were much preferred to galvanizedsteel and to steel. For steel pipe, no treatment used wasfound to be effective.

1) Waters of low hardness (20 mg/l) and alkalinity (20mg/l) with 75 mg/l total dissolved minerals, at pH 7.0:a) For improved resistance for both copper and 90-

10 copper-nickel, liquid sodium silicate (8 mg/lSiO2 ), or 5 mg/l zinc polyphosphate formulation,or 6 mg/l of zinc monobasic phosphate formula-tion, seemed preferable to the zinc silicate formu-lation at 140 and 180°F.

b) For galvanized steel, liquid sodium silicate was notevaluated at pH 7, but at pH 8.4 (8.2 mg/l SiO

2)

it was reasonably effective at 140 and 180°F.2) Waters of very low hardness (6 mg/l) and a high alka-

linity (340 mg/l) with high chloride plus sulfate (200mg/l):a) For copper and 90-10 copper-nickel, 5 mg/l zinc

polyphosphate formulation at pH 7.0 was satis-factory at 140 and 180°F.

b) For galvanized steel, liquid sodium silicate (10 to20 mg/l SiO2 ) improved the corrosion resistanceat pH 8.0 to 8.3.

3) Waters of 65 to 90 mg/l hardness, 350 mg/l alkalinity,and 180 mg/l chloride plus sulfate at pH 7.8 to 8.1:a) For copper and 90-10 copper-nickel, either 6 mg/l

zinc monobasic phosphate formulation or zinc sil-icate formulation (8 mg/l SiO2 ), except at highvelocity for copper, provided protection at 140°Fbut the 6 mg/l zinc monobasic phosphate appearedto be preferable at 180°F.

b) For galvanized steel, liquid sodium silicate (11 to22 mg/l SiO 2) at pH 8.1 was effective, and zinc sil-icate formulation (8 mg/l SiO2 ) at pH 7.8 ap-peared to be generally somewhat better than 5.6mg/l zinc monobasic phosphate formulation andliquid sodium silicate (4 mg/l SiO2 ) at 140 and180°F.

4) Waters of 150 mg/l hardness, 120 mg/l alkalinity, and300 mg/l chloride plus sulfate at pH 6.8:a) For copper and 90-10 copper-nickel, with pH ad-

justment to 8.3, improved corrosion resistance at

3 1

140°F was noted with liquid silicate (8 mg/l SiO ), until a more thorough study can be conducted to better2

or zinc silicate formulation (8 mg/l SiO2 ) at 180°F. evaluate the types of treatment.b) For galvanized steel, with pH adjustment to 7.8,

liquid sodium silicate (10 mg/l SiO2 ) providedgreater corrosion resistance at 140°F than that ob- Validation of Performance

served at pH 8.3 with 7 mg/l as SiO 2 or zinc sili- For evaluating the resistance of metal piping to corrosioncate formulation (8 mg/l SiO2 ). and/or the effectiveness of treatment, in-line testing can be

5) Waters of 70 to 80 mg/l hardness, 260 mg/l alkalinity, a very useful tool to determine the effect of water qualityand 600 mg/l chloride and sulfate: and, if necessary, the treatment applied for protection.a) For copper, liquid silicate (11 to 22 mg/l SiO2 ), Water Survey personnel have developed a test specimen

and for copper and 90-10 copper-nickel, zinc sili- holder for removable pipe inserts, insulated from a metalcate formulation (8 mg/l SiO cylinder by a PVC38

2 ) at pH 8.2 appeared sleeve. The cylinder can be installedto have a slight corrosion resistance advantage over in the pipe line with PVC spacers. The internal surface isthe phosphate formulations at 140 and 180°F. not altered, and the dimensions are designed to prevent any

b) For galvanized steel, liquid sodium silicate (11 to distortion of stream line flow. A specimen holder for 1-22 mg/l SiO2 ) at pH 8.1 was preferable at 140 and inch inside diameter pipe is illustrated in figures 31 and 32.180°F to 5 mg/l zinc polyphosphate formulation, This corrosion tester, developed by the State Water Survey,and 6 mg/l zinc monobasic phosphate formulation has been adopted as an ASTM-D2688 Standard.at pH 7.7 was preferable to zinc polyphosphate The inserts are prepared by reducing the external diam-formulation at 180°F. eter to 1.125 inches in a lathe to permit an accurate weight

The assessments on treatment for 120-day periods for measurement before and after exposure. The original weightgalvanized iron were made on the basis of change in weight of the insert is approximately 100 grams. The external sur-between the original weight of the test specimens, and the face and the ends of the insert are painted with a coat ofweight recorded after removal of all loose scale and corro- epoxy primer to avoid undue corrosion of the exterior andsion products. It was not possible to remove the tight scale the end edges. After exposure, the paint is removed with anwithout removing some of the remaining zinc. The weight epoxy stripper and the specimen properly cleaned fromloss (or weight gain, indicated as a negative value) therefore scale and/or corrosion products before drying and weighingincluded the actual corrosion loss plus the hard tight scale again to determine the weight loss.from the corrosion product and the treatment. The minimum time of exposure is usually about 60

Although basically five types of treatment were used for days for the downstream insert, and 180 days for the up-the entire study, in no case were all tested for the five types stream specimen. Results may be expressed as milligramsof waters, and in no case were the levels of treatment varied per square decimeter per day or in inches per year (see pageto suggest optimum dosage. These observations therefore 20). After weighing, the specimen is sawed lengthwise andare suggestive for each type or for all types of waters. For the corrosion is evaluated by noting the number and depthany one type, the variables of original quality are limited of pits.

RESEARCH NEEDS

There is a great lack of application of research on many It is rare that any study does not require more informa-aspects of preventive maintenance related to corrosion by tion for confirmation, better data for refinement, or morewater. At the same time, research and experience have al- exact definition of the limitations. It is also rare to findready pointed to solutions of corrosion problems. Some complete solutions. Even if fundamental principles are wellsolutions lie in metallurgy and fabrication, others in effec- known and documented, research is often required to applytive coatings, and others in chemical treatment. Applica- the principles in the most effective way. Some of the re-tions of research, which generally lag many years behind search needs related to corrosion are in the following areas.the research findings, sometimes have limitations not easily No common metal or alloy is universally acceptable forrecognized. Sometimes the cost and effort are judged not all potable water for public water supplies and domesticto be worthwhile. Sometimes the corrosion is out of view uses under various conditions. Copper tubing has becomeand not readily available for inspection. But where serious the predominant metal for plumbing, where applicable, butcorrosion is recognized and there is knowledge on preven- for many waters, it has experienced notable failures. Also,tion, failure to use the best technology is inexcusable. copper is becoming very expensive and its supply short.

32

Figure 31. Insert, sleeve, and spacer of State Water Survey corrosion tester

As an alternate, thin wall stainless steel for household use cause chloride can increase corrosion rates, it would beshows some promise, but fabrication problems have yet to desirable to define more precisely the rates of increase ofbe overcome. Also, work being done on protecting gal- corrosion in various waters under conditions of highervanized pipe with a tightly adherent alumina-silicate coating chloride.should be continued. Because drinking water standards limit the concentra-

The performance of each metal and alloy is often dif- tion of lead (0.05 mg/l) and cadmium (0.01 mg/l) which isferent for waters of different water quality characteristics. an impurity in zinc, and because lead and zinc are exposedThere is a great need to clearly define the limitations of each to potable water in plumbing and appurtenances, it is im-for the various types of water under different temperature portant to know what concentrations are present at theand velocity conditions. This information is needed to de- household tap resulting from corrosion. Copper impartstermine the best combination of metals and the necessary some taste to water at 1 to 5 mg/l and stains plumbing fix-treatment. tures. Anodic stripping analytical techniques can detect very

A companion need is to determine more clearly the ef- low concentrations and can be a useful tool to evaluate thesefectiveness of various corrosion inhibitors for different corrosion products of health and esthetic significance. Ametals, alloys, and coatings – again for the various water survey of water quality characteristics in public suppliesquality characteristics. Limitations on inhibitor composi- related to concentrations of these metals as corrosion prod-tion and concentration for potable waters, related to toxic- ucts should have a high priority.ity and cost, require the best combination of materials of Corrosion often progresses undetected because it is outconstruction, and if necessary the best inhibitor. of sight in distribution systems and household piping. There

Increasing chloride concentrations are evident in many is a need to evaluate and improve upon the polarizationwater supplies. The source is usually from direct discharge technique for internal instantaneous corrosion rate measure-to streams and lakes from road salt, waste treatment efflu- ment. This is particularly important for evaluating the ef-ents, industrial wastes, and on occasion salt mining. Be- fectiveness and control of treatment with inhibitors.

33

Figure 32. Cross section of insert, spacer, and union, and assembled State Water Survey corrosion tester

REFERENCES

1 Corrosion testing bulletin. 1944. InternationalNickel Company, p. 17.

2 Speller, F. N. 1951. Corrosion, causes and preven-tion. McGraw-Hill Book Company, Inc., NewYork, Third edition, p. 420.

3 Whitman, W. G., R. P. Russell, and V. J. Altieri.1924. Effect of hydrogen-ion concentration onsubmerged corrosion of steel. Industrial andEngineering Chemistry, v. 16:665.

4 Baylis, J. R. 1926. Factors other than oxygen in-fluencing the corrosion of iron pipes. Industrialand Engineering Chemistry, v. 18:370.

5 Larson, T. E. 1960. Loss in pipeline carrying capac-

ity due to corrosion and tuberculation. JournalAmerican Water Works Association, v. 52:1263.

6 Larson, T. E., and R. V. Skold. 1958. Current re-search on corrosion and tuberculation of cast iron.Journal American Water Works Association, v. 50:1429.

7 Baylis, J. R. 1953. Effect of water velocity on thetuberculation of cast iron pipe. Pure Water, v. 5 :58.

8 McCauley, R. F. 1960. Use of polyphosphates fordeveloping calcite protective coatings. JournalAmerican Water Works Association, v. 52:721.

9 Eliassen, R., C. Pereda, A. J. Romeo, and R. T.Skrinde. 1956. Effects of pH and velocity on

34

corrosion of steel water pipes. Journal AmericanWater Works Association, v. 48:1005.

10 Larson, T. E., J. C. Guillou, and L. M. Henley. 1960.Circulation of water in the Hammond distributionsystem. Journal American Water Works Associa-tion, v. 52:1059.

11 Langelier, W. F. 1936. The analytic control of anti-corrosion water treatment. Journal American Wa-ter Works Association, v. 28:1500.

12 Ryznar, J. W. 1944. A new index for determiningamount of calcium carbonate scale by water.Journal American Water Works Association, v. 36:472.

13 McCauley, R. F. 1960. Controlled deposition ofprotective calcite coatings in water mains. Jour-nal American Water Works Association, v. 52:1386.

14 Caldwell, D. H., and W. B. Lawrence. 1953. Watersoftening and conditioning problems — solution by

15 Dye, J. F. 1958. Correlation of the two principal

16 Larson, T. E., and F. W. Sollo, Jr. 1967. Loss in

17 Larson, T. E., R. V. Skold, and E. Savinelli. 1956.

18 Skold, R. V., and T. E. Larson. 1957. Measurement

19 Larson, T. E., and R. M. King. 1954. Corrosion by

20 Borgman, C. S. 1937. Initial corrosion rate of mild

21 Evans, V. R. 1927. Practical problems of corrosion.

22 Mears, R. B., and V. R. Evans. 1935. The probabil-

23 Fraser, O. B. J., O. E. Ackerman, and J. M. Sands.1927. Controllable variables in the quantitativestudy of submerged corrosion of metals. Indus-trial and Engineering Chemistry, v. 19:332.

24

25

Larson, T. E. 1955. Report on loss in carryingcapacity of water mains. Journal American WaterWorks Association, v. 47:1061.

Williams, D. B. 1953. Dechlorination linked tocorrosion in water distribution systems. Water &Sewage Works, v. 100:3.

chemical equilibrium methods. Industrial andEngineering Chemistry, v. 45:535.

methods of calculating the three kinds of alka-linity. Journal American Water Works Associa-tion, v. 50:801.

water main carrying capacity. Journal AmericanWater Works Association, v. 59:1565.

Tuberculation of tar coated cast iron in GreatLakes water. Journal American Water WorksAssociation, v. 48:1274.

of the instantaneous corrosion rate by means ofpolarization data. Corrosion, v. 13:139t.

water at low velocity. Journal American WaterWorks Association, v. 46:1.

steel. Industrial and Engineering Chemistry, v. 29:815.

Journal Society of Chemical Industry, v. 46:347.

ity of corrosion. Transactions Faraday Society,v. 31:529.

Larson, T. E., and R. V. Skold. 1958. Laboratorystudies relating mineral quality of water to corro-sion of steel and cast iron. Corrosion, v. 14:285t.

Larson, T. E., and R. V. Skold. 1957. Corrosionand tuberculation of cast iron. Journal AmericanWater Works Association, v. 49:1294.

Larson, T. E., and R. V. Skold. 1958. Current re-search on corrosion and tuberculation of cast iron.Journal American Water Works Association, v. 50:1429.

26

27

28

29

30

Uusitalo, E., and J. Heinaneu. 1962. Corrosion ofsteel in softwaters. Corrosion Science, v. 2:281.

Stumm, W. 1959. Estimating corrosion rates in wa-ter. Industrial and Engineering Chemistry, v. 51:1487.

31

32

Pourbaix, M. 1966. Atlas of electrochemical equi-libria in aqueous solutions. Permagon Press, Lon-don, p. 411.

Cox, G. L. 1931. Effect of temperature on Corro-sion of zinc. Industrial and Engineering Chem-istry, v. 23:902.

33

34

35

Roetheli, B. E., G. L. Cox, and W. B. Littreal. 1932.Corrosion of zinc in oxygenated aqueous solutions.Industrial and Engineering Chemistry, v. 3:73.

Second corrosion study of pipe exposed to domesticwaters. 1970. Committee report. MaterialsProtection and Performance, v. 9:34.

Lane, R. W., C. H. Neff, and S. W. Schilsky. 1971.Silicate treatment to inhibit corrosion of hot, po-table water supply systems, phase 1. Technicalreport AFWL-TR-71-58. National Technical In-formation Service, U. S. Department of Commerce.

Lane, R. W., C. H. Neff, and S. W. Schilsky. 1973.Silicate treatment to inhibit corrosion of hot, po-table water systems. Technical report AFWL-TR-73-84. National Technical Information Service,U. S. Department of Commerce.

Lane, R. W., T. E. Larson, C. H. Neff, and S. W. Schil-sky. 1973. Silicate treatment inhibits corrosionof galvanized steel and copper alloys. MaterialsProtection and Performance, v. 12:32.

Lane, R. W., and C. H. Neff. 1969. Materials selec-tion for piping in chemically treated water systems.Materials Protection and Performance, v. 8:27.

36

37

38

35

Appendix A. Corrosion in Water Wells and Pumps

Corrosion in water wells is particularly difficult to assessfor both cause and degree, because there is no possibilityfor actual observation and measurement while in service.Pumps in deep wells at deep settings are of particular con-cern because of the cost and time for removal and replace-ment. Also, the purchase cost of column pipe, shaft, andshaft tube is often greater than the cost of the turbine itself.Unless a regular schedule for removing the pump to assessdamage is practiced, the pump may fail during a criticalpumping period. If the scheduled interval is too short, andno damage is noted, the cost of removal of the pump andits return to the well can be an excessive burden on themaintenance budget.

A variety of metals can be used to construct verticalturbine pumps. Bronze, cast iron, steel, and stainless steelare often used together for an acceptable period of usefullife with no serious problems at all. However, copper bear-ing metals with a metallic connection in the water to steelor cast iron always provides a potential for galvanic actionwhere the copper bearing alloy is cathodic to the anodicferrous metal. Galvanic action is accelerated by highlymineralized water, and by close physical proximity of dis-similar metals.

Within the turbine pump, the shaft to which the bronzeimpellers are attached is usually stainless steel. These twoalloys are relatively close in the galvanic series (see table 1,page 2) and there is seldom any significant galvanic actionexcept with very highly mineralized water. If the shaftwere made of carbon steel, severe corrosion of the exposedshaft could take place between the bronze bearing and thebronze impeller.

The bowls surrounding the impellers and the shaft areusually made of cast iron, and often have an acceptableservice life with bronze impellers because the bowls are notadjacent to the shaft and the impellers. However, the in-side of the bowls is subject to corrosion particularly whenexposed to waters with an appreciable proportion of non-carbonate minerals or waters containing high concentrationsof carbon dioxide. The high turbulent velocities generatedby the impellers prevent adherence of protective corrosionproducts. Severe problems with cast iron bowls are over-come by substituting all bronze bowls.

If the bowls are cast iron, current can flow from the castiron to impellers. Each stage of the pump is designed to pre-vent back flow to the underside of the impeller. This is ac-complished by providing a skirt at the underside of the im-pellers leaving very little clearance between the bronze skirtand the surrounding cast iron. This is the ‘heart’ of thepump where slippage cannot be permitted without impairingthe efficiency. Close clearance here provides minimum re-sistance to current flow, and therefore the greatest corro-sion rate of the cast iron. In many cases manufacturers

36

circumvent this by inserting a permanent or replaceablebronze wear ring attached to the cast iron surrounding theimpeller skirt, thereby eliminating accelerated corrosion atthis vital point by transferring the current to less vital areasof the bowls and reducing the concentrated attack.

Some attempts have been made to construct the turbinewith all ferrous metals rather than dissimilar metals – steelshaft, cast iron bowls, and cast iron impellers. None of thesehave been successful.

Below the bowls there is often a tailpipe which may be1 to 10 feet long in order to guide the water to the bowls.In some installations a slotted strainer is attached to the bot-tom of the tailpipe. The need for a long tailpipe is unclear,and the need for a strainer is even less clear since both offera restriction to flow and tend to release carbon dioxide andother gases naturally in the water to escape and cause corro-sion and other difficulties.

Even with the best of design of the bowl sections, othercorrosion problems may take place in the column pipe as-sembly to transport the water to the surface. The assemblymay include an oil lubricated shaft with bearings and an oilenclosure tube, or a water lubricated shaft with bronzebearing retainers inserted at intervals between column pipesections. The oil tube is similarly supported by ‘spiders’ atintervals.

For water lubricated pumps, the bearing for the shaft isusually a type of synthetic rubber, and the bearing retaineris brass or bronze. With a stainless steel shaft few if anyproblems occur, but carbon steel shafts have been observedto corrode immediately opposite the upper and lower endsof the retainer for the bearing.

Some installations apparently permit a series of galvaniccells within the column pipe assembly and the adjoiningturbine unit which enhance each other and increase therate of corrosion. This is similar to short circuiting the openterminals of two batteries which have been connected inseries. In figure A, a self-imposed current can flow fromthe cast iron bowls to the impellers and be conducted upthe shaft to flow to the bearing retainers or the bearing (ifthe water is permitted to enter the lower end of the oilenclosure tube), and return to the bowls by way of con-duction through the column pipe or the oil enclosure tube.

For either type of lubrication, steel column pipes havebeen observed many times to corrode more or less severelywithin about 2 feet above the discharge bowl of the turbineunit. For oil lubricated pumps similar action takes place atthe outside of the oil tube. To a progressively lesser extentthis action on the column pipe and on the oil tube occursjust above the bearing retainer and the spiders. Each ofthese points of attack is located where there has been a re-striction and a sudden change from velocity head to pres-sure head, the maximum pressure head being just above the

Figure A. Cross section showing areas within deep well pumps subject to corrosion attacks

turbine unit. (Similar pitting has also been noted within theturbine unit on the high pressure side of the guide vanesabove the impellers.)

It can be assumed that gas bubbles are released from thewater at low pressure points on the underside of the im-pellers, and at constrictions and eddies created by the ex-tremely rapid changes in velocity, direction, and pressurewithin the turbine unit. As the water passes through eachsuccessive bowl, the gas bubbles and the water are sub-jected to increased pressure, and upon leaving the dischargebowl, the larger diameter of the column pipe converts thevelocity head to additional pressure head. This pressuretherefore provides a greater potential for solution of the

gas bubbles. Normally, the gases are nitrogen, carbon diox-ide, and water vapor. As the carbon dioxide dissolves thewater becomes excessively high in carbonic acid at the inter-face with the bubbles, because the acid diffusion rate in wa-ter is very low. As the acidic surfaces of the bubbles passover the column pipe surface, corrosion is enhanced. Asthe water passes up the column pipe the bubbles eventuallydissolve completely, and the interface is eliminated, there-by reducing the corrosion.

This type of attack has also been observed on the highpressure side of booster pumps, in valves, and elbows. Thepresence of dissolved oxygen or hydrogen sulfide can accel-erate this type of corrosion.

37

AP

PE

ND

IX B

83

REPRINTED FROM AND COPYRIGHTED AS A PART OF

JO U R N A L AM E R I C A N WA T E R WO R K S ASSOCIATIONVol. 58, No. 10, October, 1966

Printed in U. S. A.

Deterioration of Water Qualityin Distribution Systems

T. E. LarsonA paper presented on Mar. 16, 1964, at the Missouri Section Confer-ence on Corrosion, Columbia, Mo., and on Sep. 16, 1964, at theChesapeake Section Meeting, Ocean City, Md., by T. E. Larson,Head, Chemistry Sec., State Water Survey, Urbana, Ill.

MOST annoying to all water utilitypersonnel, as well as to the con-

sumer, is deterioration of water qualitythat occurs between the plant and thehousehold tap. Such deteriorationtakes place after plant treatment byacceptable practices at reasonable, ifnot insignificant, cost to the consumer.This is somewhat akin to serving himperfectly good drinking water in anunclean container.

The type of quality deterioration isnot of concern to the consumer; any-thing less than a palatable, safe, andaesthetically pleasing water is neitherexpected nor accepted if he is payingthe margin between the cost of waterfor fire protection and a quality prod-uct for household use. On the otherhand, if he is not paying the necessarymargin for a quality product, the con-sequences are of his own making.

Types of Deterioration

For analysis of quality deterioration,however, the type of deterioration ismost important. For obvious reasons,the present discussion must be limitedto water supply sources of reasonablequality. This can immediately elimi-nate a considerable number of sources.Moreover, certain types of deteriora-tions, such as red or black water result-

ing from oxidation by oxygen or chlo-rine of the natural iron or manganesein the water delivered to the systemand deteriorations caused by bad crossconnections and inferior householdwater-using equipment, may be tem-porarily put out of consideration.Other types of deterioration may resultfrom historical or external causes, butit would be difficult to eliminate thesewithout seriously limiting the objec-tive of this discussion.

At household taps or at hydrants, theobservable effects of deterioration arenumerous. The water may be rusty,or, although it may be clear whendrawn, subsequent reaction of solubleiron with dissolved oxygen from the airmay cause it to become rusty. It maybe turbid, red, white, or black. It maybe temporarily milky white from dis-solved gases. It may have a wide vari-ety of odors. The water may be muchharder than that entering the system.It may cause blue stains on bath fix-tures or turn silverware black. It maycontain white or red filamentous par-ticles, or, at times, granular particles.Not infrequently, with or without ob-vious deterioration in quality, theremay be a lack of pressure not only atthe tap but also, and most dangerously,at the fire hydrant.

1308

State of the Art

T. E. LARSON Jour. A W W A

Disorders of Age

Manganese dioxide deposits havebeen responsible for loss in carryingcapacity, and for black and even purplewater, after excessive polyphosphatetreatment with chlorination.

Pipe tuberculated by corrosion alsosuffers from loss in carrying capacityand causes a loss in chlorine residual.In every case, it is difficult to protect thepipe by calcium carbonate treatment orany other means of preventing red wa-ter. The degree of difficulty dependson velocity of flow and degree of dirti-

Given the present state of knowledge,there is little reason for any new wa-ter distribution system to suffer fromother than the rarer types of qualitydeterioration. The major problems ofdeterioration today should be almostwholly found in old distribution sys-tems with any of a number of failings,in some cases abetted by inadequatelyequipped or operated water plants. ness or tuberculation. Mechanical

Although many developments are inthe making and research studies prom-ise further improvements not only intreatment practices but also in con-struction materials and distributionsystem design, knowledge of treatmenthas advanced far in the past quartercentury. Langelier’s saturation index,for example, has extended greatly theapplicability of calcium carbonate sta-bility as a means of corrosion control.Cement-lined cast-iron and steel pipefor corrosion control is now availableat little or no greater cost than coal-tar-lined pipe. Other pipe materials arealso available, and satisfactory meanshave been developed for reliningcleaned pile. New jointing materialshave obviated the need for jute pack-ings, which have been a source of muchtrouble, and new trenching machinescan produce cleaner and quicker in-stallations. Chlorination in its variousforms is better understood. Use of ac-tivated carbon for reduction of tasteand odor and incidental organics is notunusual. Better means of recarbon-ation for pH control are available.Coagulation theory and the use of co-agulant aids is much better understood,so that a better quality water is moreuniformly provided, even though thereis much more to learn and to he ap-plied. Other improvements, too nu-merous to mention, are also a part ofmodern practice.

Slime growths of zoogloeal massescan he so luxuriant that, without ade-quate water circulation and chlorina-tion, chlorine residuals are lost, carry-ing capacity is lost, and taste and odorproblems develop.

Water systems are more likely to belargely old than new, however. Adistribution system designed manyyears ago often suffers from the pres-ence of mud, alum, iron oxide, man-ganese dioxide. slime deposits, or tu-berculation resulting from corrosion ofpoorly protected pipe. Mud may re-sult in dirty or turbid water, loss inchlorine residual, growth of organisms,taste and odors, corrosion, or loss incarrying capacity.

Residual alum deposits, often withthe microorganisms and turbidity sup-posed to have been removed duringfiltration, may cause loss in carryingcapacity and increase corrosion andodor problems.

Iron oxide, deposited before iron re-moval was practiced, may require yearsbefore it can be removed, even with theaid of frequent intensive flushing.Such deposits can also permit thegrowth and eventual decay of iron bac-teria, which entails red water andodor problems.

1307

AP

PE

ND

IX B

(Continued)

39

cleaning of tar-coated cast-iron piperarely, if ever, is satisfactory in all re-spects for even a temporary period,unless a corrosion-resistant lining isapplied.

Although these problems are termedas disorders of age, unfortunately theycan and do exist in new systems andnew additions to old systems—not be-cause the state of the art has not im-proved, but often because of failure toapply the newer improvements in the

1309Oct. 1966 QUALITY DETERIORATION 1310 T. E. LARSON Jour. A W W A

of what is being deteriorated by what.Consider a water distribution systemfor a municipality of about 50,000 pop-ulation, delivering 5,000,000 gal (20,-000 tons) per day to some 14,000 serv-ices under a pressure of 50–60 psithrough some 100 mi of pipe buried ata minimum depth of 6 ft. The systemconsists largely of approximately 25,-000 sections of 16–20-ft pipelengths andan equivalent number of joints, withsome 1,500 elbows and tees, 1,200

TABLE 1

Amounts of Foreign Constituents in Hypothetic System

Tot. Amt. in Tot. Amt. DeliveredConen.— mg/l System at Any Daily Through

I t e m (Unless Otherwise Time— lb* System— lb †Noted) (Unless Otherwise (Unless Otherwise

Noted) Noted)

Minerals 500 2,500 21,000Iron 0.3 1.5 12.6Manganese 0.05 0.25 2.1Chlorine-ammonia compound residuals 0.3 1.5 12.6Phenol 0.001 0.005 0.042Acetic acid 5 25 210Nitrogen 16 80 670Oxygen 8 40 335Methane 1 5 42Excess CaCO3 1-10 5-50 42-420Suspended matter 0.1 0.5 4.2Ammonia 1 5 42Deuterium oxide (HDO) 156 780 6,600Radium 226 3 µµ C/1 6.8 g 57 gStrontium 90 10 µµ C/1 22.5 g 190 gColiform organisms 1 per liter 2,250,000 units 19×10 6 unitsTot. organisms 100 per liter 220,000,000 units 1,900×10 6 units

*Total amount of water is 5,000,000 1b (600,000 gal. 2,200,000 liters).†Total amount of water is 42,000,000 1b (5 mgd, 19,000,000 liters).

art. Whether this be the result of valves, and 900 attached fire hydrants.negligence or penny-wise, pound-fool- The system is designed to withstandish decisions is not the object of thisdiscussion. Rather. the object is tocall attention to the nature of distribu-tion systems, the problems that canexist, and their causes.

Hypothetic System

Before discussing deterioration fur-ther, it is important to have some idea

external as well as internal shockstresses of several hundred pounds persquare inch. In addition, water storageand pressure regulation would requireone or two reservoirs, elevated orground surface, with pumping facilities.

Materials in contact with the watermay vary from metallic to mineral toorganic; each can have coatings of

various compositions, thicknesses, de-grees of adherence, and porosities, ornatural surface conditions peculiar tothe composition and the method ofmanufacture.

The system is designed for minimumstandards for fire flows as well as ade-quate pressures to deliver water fordomestic, commercial, and industrialuses. The normal velocity of flow mayrange 0–6 fps on a continuous or inter-mittent basis. It is not unusual that,in certain sections of the system, thewater may remain between the point ofentry and the point of withdrawal forseveral days.

The temperature of the water mayrange seasonally from near freezing toas much as 80°F. The temperaturefrom the point of entry to the point ofwithdrawal may also vary several de-grees, depending upon the season andthe time the water remains in the sys-tem.

The volume of water in the system,exclusive of that in storage, is probablyabout 600,000 gal. This water maycontain as much as 500 mg/l of as-sorted minerals, largely as ions, 1 coli-form organism per liter, 100 other or-ganisms per liter, 0.3 mg/l chlorine (aschlorine-ammonia compound residual),1 µg/l phenol, 0.3 mg/l iron, 0.05 mg/lmanganese, as well as a host of un-knowns. The concentration of suchconstituents in the system at any timeand the total amount transmitted perday are given in Table 1.

Foreign ComponentsWhether they are particulate suspen-

sions or dissolved as a homogeneoussolution, these foreign substances oc-cupy space in the water. In solution,they are surrounded by water, beingbonded with the water molecules,trapped in cages of self-bonded water

molecules, or suspended by virtue oflow specific gravity and possible hy-dration.

Relative sizes of the components ofwater and foreign substances are givenin Table 2. These indicate an orderof relative spatial magnitudes. Thevolumes occupied may, generally, be as-sumed to be proportional to the cube ofthe components’ diameters; hydrationof precipitated hydroxides further in-creases the volume.

Equilibrium

Although most minerals are rela-tively stable in relation to each other,their ions are electrically charged andusually separated as the result of the

high dielectric constant of water. Thus,these ions are in a state of constant at-traction and repulsion amongst them-selves within a liquid whose moleculesalso possess discrete positive and nega-tive charges.

It is convenient to think that associ-ations between ions and undissociatedmolecules are in equilibrium. Thisdoes not necessarily mean that the un-dissociated molecules are stable.Rather, dissociation of specific pairs ofassociated ions takes place and othersbecome associated. This equilibriumrelates more to the probability that a

TABLE 2

Relative Sizes of Components of Water andForeign Substances

Component Particle Size—µ

H2 O(1) 0.0003H2 O(g), O 2, CO2 , CH 4 , N2 0.0004–0.0005Ions 0.0002–0.0004+Virus 0.01–0.5Turbidity 0.5–10Bacteria 0.3–30Fine sand 20–200

Oct.1966 QUALITY DETERIORATION 1311 1312 T. E. LARSON Jour. A W W A

proportion of specific pairs of associ-ated ions exist relative to the disas-sociated ions of the same species. Itshould be considered a dynamic statesensitive to such physical factors astemperature and to its elements’ as-sociation with other species of ions.

Although equilibrium is a conditionwhich, for practical purposes, may besaid to exist, it should be recognizedthat it is a condition approached at anever decreasing rate as the criteriafor fulfillment are approached.

The number of water entities, if notassociated with each other, would totalabout 5,000 times the various otherentities. The water entities in them-selves, particularly through their self-association, have a profound effect oninteractions between the ionic speciesand on their rate of movement by dif-fusion. Thus, mechanical movement ofthe water body as a whole—as practicedin coagulation—promotes interreactionsof the entities and enhances the ap-proach to equilibrium.

ReactionsA distribution system, then, carries

a liquid that is inherently dynamic, andcontinuing changes are in turn stimu-lated by the movement of the waterthrough the system. Aside from sim-ple chemical equilibria in solution, itis possible that new particulate mattermay he formed from these reactions assolubility equilibria are approached.There can also be selective adsorptionon inorganic or biologic particulatematter. There can be enzyme reac-tions within organisms and consequentcontribution of new chemical ingredi-ents to the water. The organisms maymultiply under favorable conditions, orthey may be starved to dormancy or todeath and subsequent decay.

It is probable that all reactions ofsoluble entities with particulate mat-ter are basically surface reactions, gov-erned by diffusion and adsorption, orsubsequent absorption, and thereforeare dependent upon the dynamics of thesystem.

Of no less significance is the retain-ing interior surface of the distributionsystem, which constrains the waterfrom entry to exit. Here, too, aremany possibilities for surface reactions.likewise governed by diffusion and ad-sorption and dependent upon the dy-namics of the system. In this hypo-thetic distribution system, with perhaps1,300,000 sq ft of pipe exposed to wa-ter,¹ a gallon of water may be exposedto as much as 80,000 sq ft of pipe, or asingle square foot of pipe may be ex-posed to 1 mgd.

Influent QualityToday, when improved practices are

changing treatment from an art to afledgling science, quality failure is anespecially important index of inade-quacy of knowledge regarding treat-ment. Whether failures should be at-tributed to treatment practice, design,or condition of the system is not al-ways apparent. Certainly, any or all ofthese factors could be causative; muchremains to be known about all of them.

It is possible for coliform and otherorganisms to withstand treatment andpass into the distribution system.Treatment to condition the wateragainst corrosion of poorly protectedmetals is still an uncertain science,complicated by changing conditions ofusage and the various conditions of thedifferent segments of the system.

Biologic DisordersSo, aside from disorders of age, there

can also be disorders of what might be

called quality control. These fall intotwo broad categories : the microbiologicor biologic and the chemical.

Iron bacteria. The presence of thesemay be considered a disorder of age.Whether they are Crenothrix, Gallio-nella, or others; such as Leptothrix,iron must be present. It has been pro-posed recently that Leptothrix is aform of Sphaerotilus and can accumu-late either iron or manganese.² In anyevent, iron bacteria should not be pres-ent in properly treated water in a cleandistribution system.

Coliform organisms. The coliformorganism is an indicator of pollution.Given a raw-water supply of reason-able quality with proper coagulation,chlorination at an adequate retentiontime, and a prescribed chlorine resid-ual entering the distribution system,there should be no coliform organismsin the distribution system. Suchtreatment should also preclude thepresence of typhoid organisms. In theabsence of residual chlorine in a dirtysystem, however, coliform organismscan exist.

Other microorganisms. It is doubt-ful whether anyone will ever be ableto enumerate all of the organisms thatmay be found in distribution systems.Nevertheless, a few have been found toproduce deleterious effects on distri-bution system water quality. Whensuch organisms are prolific enough tocause a slick, slippery coating or avisible slime on the pipe wall, a meas-urable chlorine demand should be pro-duced by oxidation of the organismsor of their metabolic products. Thedemand may be so great that it becomesimpossible to obtain chlorine residualsin the remote portions of the systemexcept during high-velocity flushing.

Nitrosomonas. Nitrosomonas can de-velop as a slime on the walls of pipe.

This happens in the presence of am-monia and dissolved oxygen. Theslime can slough off and appear at hy-drants or at the household tap. Underanaerobic conditions, putrefaction canresult in taste and odor problems. Inwaters containing high ammonia con-centrations, breakpoint chlorination isimpractical; for small systems, thepumping capacity and elevated storageare not sufficient to provide adequateflushing. The use of copper sulfatewith chlorine-ammonia compoundsserves to keep the growth in check.Lime softening to raise the pH alsoappears to prevent such growths.Methanomonas also reportedly pro-duces slime, in the presence of methaneand dissolved oxygen. There is, how-ever, little documentary information onthis subject.

Sulfate-reducing bacteria. These util-ize the sulfate ion to form hydrogensulfide under anaerobic conditions,Such conditions may occur at pipejoints, particularly in areas of low de-mand and low-flow velocities. evenwhen dissolved oxygen is present in thebody of the water flow. Such anaerobicconditions may prevail in the cores oftubercles and under any debris thatmay be in the system. The extent towhich corrosion and tuberculation isthereby enhanced has yet to be evalu-ated. The resultant problems aretastes and odors, staining of silverwareby hydrogen sulfide, and possible ac-celeration of corrosion and tubercula-tion.

Sporovibrio desulfuricans. This andother such sulfate-reducing organismsthrive better at moderate temperaturesand in neutral or slightly acid condi-tions. In surface water, these tem-peratures prevail in all seasons exceptwinter, and the low pH prevails intubercles and debris under anaerobic

AP

PE

ND

IX B

(Co

ntin

ued

)

40

Oct. 1966 QUALITY DETERIORATION 1313

conditions. Chlorine can destroy theorganisms, but it is virtually impossiblefor chlorine to penetrate tubercles anddebris. These organisms can remaindormant for long periods even in thepresence of dissolved oxygen. Theymust have some organic matter to growluxuriantly; this could either be pro-vided by dead cells in tubercles or ab-sorbed from the distribution systemwater.

Beggiatoa. Sulfide-oxidizing bac-teria thrive as a slime in the presenceof hydrogen sulfide and dissolved oxy-gen, to form sulfur. The presence ofsuch organisms might also be con-sidered as a disorder of age, becauseadequately treated water in a cleandistribution system should have no hy-drogen sulfide.

Actinomycetes. These have been re-ported as the source of taste and odorproblems in heavily polluted water sup-plies and when putrefaction occurs onthe banks of streams during low-flowconditions. In some instances, it hasbeen reported that water which is odor-free after activated-carbon treatmentbecomes foul on passing through thedistribution system, even in the pres-ence of chlorine. The records are notclear on this problem, and much ap-pears to be conjecture.

Vegetative organisms. Algae, Creno-thrix, Sphaerotilus, and other vegeta-tive organisms have also been reportedin water mains. Crustacea such aswater fleas, mollusks such as snails,nematodes, and other animals are notunusual. Whether all of these shouldbe considered to cause deteriorationof water quality in the mains orwhether they are transient inhabitantsis an open question. Certainly, theirpresence indicates an inferior waterquality. It appears that water lice(Asellus aquaticus) may survive and

multiply in the system for long periodsof time. Their removal can be stimu-lated by the use of phrethin and flush-ing.³

CorrosionDeterioration of water quality

through changes in mineral contentcan often take place as the result of“dirty” pipe, use of pipe poorly pro-tected against corrosion, from treat-ment inadequate to reduce or preventcorrosion, or from circulation inade-quate to aid in protection against cor-rosion. An absence of red water atthe tap or hydrant is no assurance thatthe system is not suffering from corro-sion or tuberculation. The greatestnumber of problems seem to be due toa combination of poorly protected pipe,poor circulation, and inadequate treat-ment.

Most corrosion problems are associ-ated with relatively soft waters contain-ing dissolved oxygen. The logical wayto eliminate this problem would betreatment for oxygen removal; how-ever, capital and operation expenseshave been prohibitive.

As a rule, hard water is less corro-sive than either water softened by limeor ion-exchange treatment or naturallysoft water, particularly when dissolvedoxygen is present.

A third class of treated waters alsois notorious for suffering deteriorationin the distribution system, whether suchwater has an appreciable hardness ornot. These are coagulated and filteredsurface waters. In too many of theseinstances, no doubt, quality problemsresult from the failure even to attemptto adjust the pH to a satisfactory leveland provide a water that is “stable,” orprotective. There is evidence, how-ever, that, in some cases, where the pHis not greatly different from the stable

1314 T. E. LARSON Jour. A W W A

level, few problems are experienced.In others, there is evidence that, withthe best advised pH adjustment, evento a positive index of supersaturationwith respect to calcium carbonate,problems are rampant.

There is little point in elaborating onprinciples of corrosion and its preven-tion, or on the validity of the theoryand equations related to the solubilityof calcium carbonate. There is moremerit in focusing attention on the fail-ures of sincere applications of the prin-ciple of calcium carbonate protectionfor distribution systems. They are notmerely tormenting exceptions.

TuberculationPrevious comments on dirty and

tuberculated pipe and on velocity offlow probably require further elabora-tion. In the presence of dissolved oxy-gen, active tubercles can sustain them-selves through production of hydrogenions, as ferric hydroxide is precipi-tated. Such acidity, associated with apredominance of sulfate and particu-larly chloride ions in low-alkalinity wa-ter, eventually results in an actual acidiccondition, if the buffering effect of thebicarbonate-carbon dioxide equilibriumis overcome. Thus, the proportion ofchloride and sulfate to alkalinity is afactor related to the development andcontinued growth of tubercles and tocorrosion in general.

The total alkalinity is also an inde-pendent and important factor, servingas a buffer to change in pH. Stumm 4

has recently reviewed the effect of pHon the types of alkalinity and the re-sultant effect on the buffer capacity.It has also been shown that low al-kalinity (<35 mg/1) offers little buffereffect on pH as the temperature is in-creased in hot water tanks.5 For such

waters, temperature change can re-verse a positive to a negative saturationindex.

The quantity of acidity developed byactive tubercles can seriously decreasethe pH in the area of the tubercles andnullify the designed favorable state ofCaCO3 supersaturation of the water atthe surface of the tubercle.

Similarly, mechanical cleaning ofpipe, by removing sections of coatings,increases the areas open to corrosion.Corrosion can then proceed at a rategreater than the rate of protection thatcan be expected from the water.

Flow VelocitiesHigh velocities of flow are related

either to the protection or the degrada-tion of the quality, or both. Underlaminar low flows, the lateral movementof ions and molecules from areas ofhigh to low concentration at the pipewall must be by diffusion. The ratesof diffusion under the very low concen-tration gradient are extremely slow.At increasingly turbulent flow rates,the diffusion distance through the lami-nar layer at the pipe wall is decreased,and the corroding elements (dissolvedoxygen and hydrogen ions or both) aswell as the protecting elements (cal-cium and alkalinity) approach the pipewall correspondingly more rapidly. Ifthe corroding elements predominate,corrosion is accelerated; but if the pro-tective elements predominate, protectionis generated more rapidly up to thepoint of their depletion in distance ortime. The higher velocities, of course.provide a greater quantity of corrosiveor protective elements to the pipe wallper unit of time. These elements aretherefore related to the development ofactive tubercles and their rate of corro-sion or degree of activity.

AP

PE

ND

IX B

(Co

ntin

ued

)

41

Oct 1966 QUALITY DETERIORATION 1315

Protective DepositsMention should be made of the

amount of deposit required to protectuncoated surfaces. If 1 mi of 6-inpipe were to receive a uniform depositof calcium carbonate 1 µ in thickness,the total weight of the deposit wouldbe 5.85 lb. In terms of mils (0.001in.), the weight would be 146 lb permil per mile. Thus, even with a cal-cium carbonate concentration 10 mg/1in excess of solubility, more than 1.8mil gal would have to be transmitted ifall the excess calcium carbonate weredeposited uniformly in a layer 0.001 in.thick from water on 1 mi of cleanedpipe. Such a flow in a short periodof time is impossible at dead-ends.

McCauley’s work on deposition ofcalcium carbonate as a thin transparentlayer on clean steel and sand-blastedcast-iron pipe shows that such protec-tion requires deposition be amorphousor microcrystalline rather than in thelarge, visible calcite form.6 His processinvolves an excess of calcium carbon-ate solubility of as much as 150 mg/1and the use of polyphosphate in a rela-tively small concentration (1–2 ppm)in combination with lime and soda ashat a flow rate of 4 fps for 2 hr. Thephosphate seems to inhibit depositionof large crystals, thus providing a morecontinuous protective film, and to per-mit adherence to the metal wall.

From the observation of properlyprotected specimens, the deposition ap-pears to be almost transparent. Thisis quite different from the usual “eggshell white” type coating. McCauley’scoating appears to be the type that pro-tects pinholes or holidays in tar-coatedpipe in the same manner that so-callednoncorrosive waters do. In many suchcases, the pipe shows few points of

corrosion and the tubercles are verysmall and inactive.

Such noncorrosive waters are nottreated with polyphosphates, however.It may therefore be deduced that thesewaters contain a natural ingredient thatprovides the same protective effect asthe polyphosphate in McCauley’s pro-cedure. This natural ingredient mightalso be considered more versatile, forit does not appear to be as dependenton velocity and cleanliness of pipe sur-face. This deduction harmonizes withthe observation that treated waters aredifficult to adjust for corrosion control.Inasmuch as treated waters are de-signed to remove undesirable ingredi-ents from the water, it is probable thatsuch treatment also removes a desira-ble unknown ingredient.

Silica and ColorTwo natural components that may

possess this inhibiting property aresilica and color. Too little is knownabout either. Silica could be in a poly-molecular form and color may also havesimilar combining properties. Suchproperties would be associated withdisruption of crystal growth, the bind-ing of colloidal microstructures to eachother as a film, and the concurrentbinding of these to the metal surface orits oxide. Or, again, they may preventhydration of iron oxide as it is formedby corrosion.

This speculation is supported by ob-servations by Turner and Campbell atthe Non-Ferrous Metals Institute inEngland, where synthetic watersformed nonprotective calcium carbon-ate and zinc carbonate on galvanizediron, but natural water did not formsuch carbonates and did not corrode.When a natural organic extract, in-cluding color, was added to the syn-

1316 T. E. LARSON Jour. A W W A

thetic water, protection with a trans-parent deposit was achieved.

ConclusionA distribution system is a sensitive,

dynamic, living individual with its ownpeculiar characteristics, not just a net-work of tubes joined together. Al-though it may not always seem so,planning is needed to permit necessaryquantities of water to reach the rightplaces at the times they are needed.And again, although from this discus-sion it may not appear so, there isusually a plan to provide a qualityproduct at the point of delivery.

Unfortunately, these plans often haveproved weak and their aims have beeninterfered with by inadequate attentionand control in design, construction. use,or maintenance, as well as lack of rec-ognition of quality factors and of thenecessary treatment for maintaining anondegradable quality. The best at-tention must be given at all times toeach link in this chain of endeavors,and to the coordination of all phases ofsupervision, including that essentialcalled management, to achieve the de-sired objective: delivery of qualityproduct to the consumer.

Perhaps most significant is the recog-nition that quality design, quality con-struction, quality protective measures,and quality treatment cannot be ob-tained from low-quality materials andpersonnel or an “or equal” that is oftenless than equal. Quality cannot be ob-tained at minimum cost. The distribu-tion system represents about two-thirdsof a water utility investment. Thecost of the best quality pipe and other

materials is only a fraction of the costof installation or replacement.

There is no question that watertreatment practices vary and that someare more difficult to contol than othersbecause of the raw material, the sourceof supply. The mineral quality of somesupplies, given the present state of theart, defies such treatment as wouldprevent all corrosion. Treatment canbe adequate to provide an excellentproduct for the consumer at the plantoutlet, but inadequate to overcome in-ferior protective measures in the dis-tribution system. Therefore, new addi-tions to the system must be of the bestmaterial. Old segments of the systemthat have proved to be inferior must berehabilitated. None of these areachieved without substantial additionalcost.

References1 . H A N E Y, P. D. Consulting Engineer

N e e d s . J o u r . A W W A , 5 3 : 6 9 2 ( J u n .1961).

2. W OLFE , R. S. Cultivation, Morphology,and Classification of the Iron Bacteria.Jour. AWWA, 50:1241 (Sep. 1958).

3. T U R N E R , M. E. D. Asellus aquaticus i na Publ ic Water Supply Dis t r ibut ionS y s t e m . P r o c . S o c . W a t e r T r e a t .Exam., 5:1:67 (1956).

4. WE B E R, W. W., JR . & ST U M M, WE R N E R.Mechanism of Hydrogen Ion Bufferingi n N a t u r a l W a t e r s . J o u r . A W W A ,55:1553 (Dec. 1963).

5. L ARSON , T. E. The Ideal Lime-SoftenedWater . Jour . AWWA, 43:649 (Aug.1951).

6 . M CCAULEY , R. F. Use of Polyphosphatefor Developing Protective Calcite Coat-ings. Jour. AWWA, 52:721 (Jun.1960).

7. T U R N E R, M. E. D., & CA M P B E L L, HPrivate Communication.

AP

PE

ND

IX B

(Co

ntin

ued

)

42

7

Bacteria, Corrosion and Red Water

By T. E. Larson

MANY investigators have devoted years to the study of corrosion.Numerous phases of this complex problem have been solved

or eliminated but many are left unsolved and are complicated by avariety of new as well as old factors—factors found to prevent cor-rosion in some cases and to increase corrosion in others.

The question is definitely one which involves chemical reactions,—reactions which are known to be possible but vary in rate and whichcan be stopped or aided by side reactions. “How fast does the re-action take place?” “How long will the reaction continue?” Thesetwo questions are answered by the controlling factors. Control therate, and corrosion may be minimized or stopped.

The present day control consists of using a protective coatingcomposed of a substance unreactive to water and oxygen. Perhapsthe most elementary protective coating is a film of hydrogen con-stituting the overvoltage phenomenon.

It is well known that pipe effectively lined with enamel, cement,asphalt, or coal tar is not subject to metallic corrosion. In everycase, with the possible exception of cement, the coating must befirmly adherent and without breaks or pinholes which have beenpointed out to be often more harmful than if the pipe were not coatedat all. Concentrated attack may take place at these points to piercethrough the metal long before general corrosion could weaken thestructure.

Softened or naturally soft water properly treated with lime soda,caustic soda, soda ash or sodium silicate has also been found toprevent metallic corrosion when a good adherent, unbroken calciumcarbonate or silicate coating has been laid down and maintained.It is essential that this coating be complete and maintained.

A paper presented at the Illinois Section meeting, Urbana, Illinois, April 21,1939, by Dr. T. E. Larson, Chemist, State Water Survey, Urbana, Illinois.

1186

VOL. 31, NO. 7] BACTERIA AND CORROSION 1187

Oxygen becomes almost a necessary evil when it is essential thataeration be the means of iron and gas removal, or taste and odorcontrol. Corrosion has been shown to take place almost in propor-tion to the dissolved oxygen content in natural water (1). On theother hand dissolved oxygen can also aid in the inhibition of corrosion(2, 3, 4) by aiding in the deposition of calcium carbonate with ferrichydroxide at the point of corrosion reaction. Here the question istied up with the velocity of flow and the various concentrations ofnegative ions.

“Red water,” the cause of numerous complaints may be from eitherof two sources. Iron may be present in the raw water in which casethere is no hope of alleviating staining and bacterial growths withoutremoval of the iron by an accepted standardized procedure. Theother source is by corrosion of the distribution mains or service lines.It may be well to state the question of the rate of corrosion in anotherway: “How much time is available for corrosion to take place?”This will be discussed later.

The plague, or should I say one of the plagues, of the water worksman’s life, consists of placing a clear, colorless, odorless water into thedistribution system at the plant and finding consumers complainingof obtaining dirty, red, smelly water from their tap. That’s dis-couraging.

Most Illinois waters are moderately hard or hard and unsoftened.The addition of chlorine or chloramine, frequently accompanied byiron removal is in most cases the only treatment used, if any. Thisis typically exemplified by the University of Illinois water. Hereonly two peculiarities may be noted. The alkalinity of 330 p.p.m.is greater than the hardness of 280 p.p.m. and ammonia nitrogen ispresent in the raw well water to the extent of 1.5 to 2.5 p.p.m. Thesodium bicarbonate presence is not considered to be a decisive factorin connection with “red water,” but the ammonia content has beenfound to be a new if not a potent factor.

Treatment at the University plant consists of iron and gas removalby coke tray aeration, sand filtration, and chlorination. As it leavesthe plant the water is iron free by test, has a chloramine content ofmore than 1.0 p.p.m. and has a dissolved oxygen content less than8 p.p.m.

AP

PE

ND

IX

C

43

1188 T. E. LARSON [ J. A. W. W. A. VOL. 31, NO. 7 ] BACTERIA AND CORROSION 1189

A rough diagram of the distribution system is shown in figure 1.Despite the condition of the water produced at the plant at the northend of the campus, frequent complaints of “red water” are receivedfrom various points throughout the system particularly at the southend two miles away. This condition could arise from two sources:first, from old deposits left in the mains before the treatment plantwas installed; second, from present corrosive conditions in the mains.A rigorous flushing schedule has alleviated much of the trouble andhas reduced complaints considerably, but the dilemma persists andat the far south “dead” end, iron water is had within twenty-fourhours after flushing. Old deposits in the mains cannot. fully accountfor occasional soluble ferrous iron content of as much as 8 p.p.m.Corrosion must at least in part be the answer here. Microscopicexamination of “red water” sludge from service lines and of depositsin the mains has in no case disclosed the presence of iron bacteria.

In general, complaints increase with increasing distance from theplant. This is in part due to the fact that the velocity of flow de-creases with increase in distance from the plant. The iron dissolvedor the decrease in dissolved oxygen per unit distance traveled by aunit volume of water near the plant is far less than that at the outerends of the distribution system. To establish a comparable time-basis, consider the time for one cubic foot of water to pass or come incontact with one square foot of iron surface.

Computation of Contact Time

As water leaves the University plant. (1,250,000 gallons per daythrough a 14-inch main) the average time of contact for one cubicfoot of water per square foot of pipe surface is 0.15 seconds. If cor-rosion does take place, the drop in dissolved oxygen is slight and theincrease in iron per unit volume of water is relatively minute. Asthe water is drawn outward away from the plant it can be seen thatthe time of contact may approach infinity as the rate of flow ap-proaches that of a “perfect dead-end.” Assuming that one-sixth ofthe water reaches the south campus at the water tower, the time ofcontact per cubic foot of water (200,000 gallons per day through a12-inch pipe) per square foot of pipe surface is 0.87 seconds or sixtimes as much as at the plant. Continuing outward to a so-called“dead” end (1000 gallons per day through a 4-inch main) the time ofcontact is 58 seconds, or 400 times as much time is available for cor-rosion to take place and iron to be picked up by the water.

AP

PE

ND

IX C

(Co

ntin

ued

)

44

FIG. 1. University of Illinois Distribution System, exclusive of lines smallerthan four-inch.

1190 T . E . L A R S O N [J. A. W. W. A.

D E K J L F I M H G

2000

.84 .70 .42

.77 .85 .84

.56 .80 .743.4 1.2 0.0

.25 .25 .030.0 0.0 .62

320. 320. 318.

.34

.46

.450.00.0

.08320.

150016005000

FI G. 2. Typical Variations as water is distributed from north to southends of campus (left to right).

V O L . 3 1 , N O . 7 ] B A C T E R I A A N D C O R R O S I O N 1191

One other factor has been found to be present to affect the rate ofcorrosion. It has been noticed that a zeolite filter using the Univer-sity water reduced the dissolved oxygen to zero. Further tests witha sand filter produced the same effect and it was also found that onpassage through the filter that the ammonia content of the waterdecreased and the nitrite content increased. H. L. White, SanitaryEngineer of the University of Illinois, has noted an increase in thenitrite concentration on passage through the filters at the plant.Nitrite interference has been noticed in tests for chlorine at outerends of the distribution system.

Accordingly seven series of analyses were made on samples of watercollected from various points in the system. The constituents de-

TABLE 1Typical Trend of Data Throughout Distribution System

(Results in parts per million)

Ammonia– N . . . . . . . 1.92

Nitrite – N . .23Nitrate – N. . .26D.O. . . . . . . . 6.3Cl2 . . . . . . . . . 1.8Fe . . . . . . . . . . . 0.0Alk. . . . . . . . . 326.6 da . 20°C.

bact. count 0

1.92 1.44 1.26 1.34 .94.32 .50 .65 .60 .85.28 .44 .70 .52 .72

6.2 5.4 3.4 4.4 3.01.4 .6 .6 .5 .180.0 0.0 0.0 0.0 0.0

326. 324. 322. 321. 320.

0 0 0 26 600

termined included ammonia, nitrites, nitrates, dissolved oxygen, freechlorine, iron, alkalinity, and 37° and 20° incubation bacterial counts.A typical series is indicated in table 1 and fig. 2.

In general the data indicate that with increasing distance from theplant, the dissolved oxygen, the ammonia, and free chlorine decreasewhile the nitrites, nitrates and six-day, 20° bacterial count increase.Figure 2 indicates the trend in values with reference to the ammoniacontent.

In order that ammonia be oxidized to nitrites or nitrates the ratioof the loss in dissolved oxygen to the loss in ammonia-nitrogen must

AP

PE

ND

IX C

(Continued)

45

1192 T . E . L A R S O N [ J. A. W. W. A.

it not for the consistency of this trend, it would be attributed toinconsistency in the raw water alkalinity. However, in no oneseries of the tests made was the alkalinity in the system greater thanor equal to that of the filter effluent. This would indicate a slightdeposition of calcium carbonate in conjunction with corrosion by

A slight but definite general decrease in alkalinity is noted. Were

coliform test results of zero, the 20° counts can be consistently shownto be over 1000 bacteria per c.c. at certain localities in the system.At least some twenty different types of bacteria other than these areknown to be present in the system. Also, many years of previousstudy by the State Water Survey on the action of nitrifying bacteriahave shown the action indicated in this study to be undoubtedlybacterial.

concentrated effort has been made in this study as yet to isolate thesebacteria since they will not grow on ordinary media and are difficultto isolate. Despite the ordinary Standard Methods 37° counts and

Ammonia oxidation to nitrites has been reported (5, 6, 7) previouslyto take place in filter beds when ammonia was used in conjunctionwith chloramine. With ideal conditions for bacterial growth it hasbeen attributed and later shown by Feben (8) to be due to bacterialaction. Cultures of these bacteria, specific in their ability to convertammonia to nitrites, have been isolated from the filter beds by Feben.These bacteria were found to be resistant to 2.0 p.p.m. chlorine. No

monia to nitrites and nitrates.

The average ratio was 6.0. Thus it may be deduced that no morethan one-third and possibly not less than one-sixth of the oxygen islost to corrosion. For the greater part it is used in converting am-

be 4.0 or 5.15 respectively. In no case was this ratio less than 4.0.

Fe(OH)2 + Ca(HCO3) 2 →CaCO3 + FeCO3 + 2H2O

goes into solution per volume of water is very slight and undetectableby analysis.

to give calcium carbonate in conjunction with ferrous and ferricoxide. Then too, any bacterial growth requires a source of carbonsuch as carbon dioxide. Calcium carbonate deposition is furtherconfirmed by analysis of deposits collected from a north campusmain in September, 1938. The analyses indicated that only one ofevery seventeen ferrous iron equivalents is replaced by calciumcarbonate. Due to the high rate of flow at this point, the iron that

V O L . 3 1 , N O . 7 ] B A C T E R I A A N D C O R R O S I O N 1193

Corrosion at this main was not noted to be particularly severe butit was definite and of two forms. Small, single and collective tuber-cles were noted having their origin at pinholes in the bituminouslining. These were hard, compact and often grown together, andwere easily picked off with a penknife. Loose, slimy, black mudappeared at pipe joints and extended to either or both sides of thejoint for various distances. Its source is quite possibly due todeposits laid down before the treatment plant was installed to removeiron and traces of hydrogen sulfide. Galvanic action may enhancecorrosion here, but the fact that a strong hydrogen sulfide odor wasgiven off from acidified samples of this mud indicates some old de-position. This mud as well as the tubercles was coated on the waterside by a hard, yellow-brown crust. The velocity of flow here wasquite high and after aeration was installed, the dissolved oxygen wasrapidly fed to the ferrous oxide on the pipe walls to produce the ferricoxide-calcium carbonate crust. This outer coating does not appearto be sufficiently impermeable to prevent corrosion but appreciableinhibition is afforded.

In connection with bacterial action it is interesting to note thatalthough the dissolved oxygen is partially used up in oxidation ofammonia instead of corroding the iron, this oxidation converts abasic ion into an acidic ion thereby producing two equivalents ofacid from one equivalent of a neutral salt. In other words, the lossof one ammonia equivalent produces two free acid equivalents.

NH4HCO3 + 30 → HNO2 + H 3 + H2CO 2O

Although the ammonia lost per volume of water is slight, the acidiccondition can be expected to be greater at the pipe walls where thebacteria are to be found. This permits two corrosion enhancingconditions. First, a decrease in the protective action of calciumcarbonate and ferric oxide is to be expected. Second, oxidizingagents are available to continue corrosion in lines where the dissolvedoxygen is depleted. The latter effect is exemplified by the conditionsin the far south “dead” end of one mile length. The dissolved oxygen

Although bacterial protoplasm contains some three and one-halftimes as much carbon as nitrogen and some of the free carbon dioxideis used up here, only a very small percentage of the ammonia loss isconverted to protoplasm. Consequently little of the carbonic acidproduced is used up in this manner.

AP

PE

ND

IX C

(Continued)

46

1194 T . E . L A R S O N [J. A. W. W. A.

entering this main is less than 1 p.p.m. and is rapidly depleted.Nevertheless the iron at the exit has been found to be as high as 8.0p.p.m. In the meantime nitrites and nitrates are reduced from 1.0to 0.1 and from 0.7 to 0.2 p.p.m. respectively (table 2 and fig. 3).The data in this figure are plotted according to distances in feet as

FIG. 3. Typical Variations in “dead” end (four-inch) after oxygen is de-pleted. (73 sec. per cu.ft. water per sq.ft. pipe surface).

measured between sampling points. Here anaerobic bacterial actionis strongly suggested.

Perhaps it should be mentioned that analyses for such low con-centrations of ingredients showing such low differences in magnitudeare sometimes questionable but the consistency and direction ofchange in these constituents nullifies any doubt of their validity.

V O L . 3 1 , N O . 7 ] B A C T E R I A A N D C O R R O S I O N 1195

SummaryA certain water may be only slightly corrosive in one part of a

system and strongly corrosive in another due to the difference in timeof contact for corrosion to take place per unit volume of water. Theconcentrations of chemical constituents affecting the rate of corrosioncan change during passage through the system to alter the rate whichcontrols corrosion. A classification of waters as corrosive and notcorrosive cannot be made without placing restrictions on all factorsaffecting the rate.

Dissolved oxygen in this water system gives rise to conditionswhich may inhibat or enhance “red water.” Inhibition is experi-

TABLE 2Typical Low Flow Data

(Results in parts per million)

S A M P L I N G P O I N T

0 1 2 3 4 5 7 8

0 . 64 0 . 6 8 0 . 6 8 0 . 5 4 0 . 5 4 0 . 5 4 0 . 5 2

1 . 1 2 0 . 7 5 0 . 6 0 0 . 3 3 0 . 2 9 0 . 2 2 0 . 2 1 0 . 0 80 . 7 2 0 . 6 3 0 . 3 9 0 . 2 5 0 . 2 0 . 1 5 0 . 2 3 0 . 1 90 . 7 0 0 . 7 6 0 . 7 2 0 . 9 8 0 . 8 8 0 . 8 4 0 . 8 4 1 . 0 4

2 . 5 4 2 . 14 1 . 7 1 1 . 5 6 1 . 3 7 1 . 2 1 1 . 2 8 1.310 . 2 0 . 1 0 . 2 0 . 2 0 . 0 0 . 0

318. 318. 320. 314. 316. 316. 31 4 . 314.0. 0. . 0 8 . 0 8 . 8 8 1 . 3 2 4 . 1 6 2 . 1 6

A m m o n i a – N .

Nitrite – N. . . .

N i t r a t e – N . . .

K j e l d a h l – N . .

( K j e l d a h l + n i -

trite + ni-

t r a t e ) – N . . .

D.O. . . . . . . . . .

Alk. . . . . . . . . .

Fe.. . . . . . . . .

enced by the formation of a more or less protective ferric oxide-calcium carbonate coating in the mains which can hinder penetrationof dissolved oxygen to the metal itself as well as prevent corrosionproducts from sloughing off. The presence of ammonia and oxygenas a source of energy for bacteria gives rise to bacterial growthsin the mains. Corrosion is enhanced by the bacterial transforma-tion of basic ammonia and oxygen to acidic and oxidizing nitritesand nitrates.

The determination of the depletion of dissolved oxygen throughoutthis system is not a true measurement of the degree of corrosion.Depletion of dissolved oxygen at the “dead” ends and service linesof the system gives rise to a condition where ferrous iron is not

AP

PE

ND

IX C

(Continued)

47

48

APPENDIX C (Concluded)

1196 T . E . L A R S O N [ J. A. W. W. A.

oxidized to insoluble ferric oxide and solution of iron to the ferrousstate takes place at the expense of the nitrite and nitrate previouslyformed—quite possibly by bacterial action.

What corrosion that does take place in the system as a whole ismagnified by the accumulation of traces of the iron in the gelatinousmasses of bacterial growths clinging to the pipe in localities where thevelocity of flow is low and sloughing off occasionally.

We are deeply indebted to Mr. White and his staff at the waterplant without whose splendid cooperation a study in this mannercould not have been accomplished.

References1. SPELLER , F. N., Corrosion, Causes and Prevention, Second Ed., p. 16 (1935).2. MCKAY, R. J., and WORTHINGTON , R., Corrosion Resistance of Metals and

Alloys. A. C. S. Monograph, No. 71, p. 67 (1936).3. VERNON, W. H. J. and WORMWELL , F. Some General Principles Involved

in the Corrosion of Water Mains and Services, Wtr. and Wtr. Eng.,40: 8 (1938).

4. BORGMANN, C. W. Treatment of Natural Waters to Prevent and ControlCorrosion. Jour. A. W. W. A., 30: 256 (1938).

5. GOEHRING, E. C. Ammonia-Chlorine Treatment at Beaver Falls and NewBrighton, Pa. Jour. A. W. W. A., 23: 9 (1931).

6. H EDGEPETH , L. L. The Practical Aspects of Taste Control. Six AnnualMichigan Conference on Water Purification.

7. HULBERT , R. Nitrites in Filtered Water Cause Troubles. Eighth AnnualMichigan Conference on Water Purification.

8. FE B E N, D. Nitrifying Bacteria in Water Supplies, Jour. A. W. W. A.27, 439 (1935).